text
stringlengths 66
377k
|
|---|
First all-in-one diagnostic tool for DNA intelligence: genome-wide inference of biogeographic ancestry, appearance, relatedness, and sex with the Identitas v1 Forensic Chip
When a forensic DNA sample cannot be associated directly with a previously genotyped reference sample by standard short tandem repeat profiling, the investigation required for identifying perpetrators, victims, or missing persons can be both costly and time consuming. Here, we describe the outcome of a collaborative study using the Identitas Version 1 (v1) Forensic Chip, the first commercially available all-in-one tool dedicated to Electronic supplementary material The online version of this article (the concept of developing intelligence leads based on DNA. The chip allows parallel interrogation of 201,173 genomewide autosomal, X-chromosomal, Y-chromosomal, and mitochondrial single nucleotide polymorphisms for inference of biogeographic ancestry, appearance, relatedness, and sex. The first assessment of the chip's performance was carried out on 3,196 blinded DNA samples of varying quantities and qualities, covering a wide range of biogeographic origin and eye/hair coloration as well as variation in relatedness and sex. Overall, 95 % of the samples (N 0 3,034) passed quality checks with an overall genotype call rate >90 % on variable numbers of available recorded trait information. Predictions of sex, direct match, and first to third degree relatedness were highly accurate. Chip-based predictions of biparental continental ancestry were on average~94 % correct (further support provided by separately inferred patrilineal and matrilineal ancestry). Predictions of eye color were 85 % correct for brown and 70 % correct for blue eyes, and predictions of hair color were 72 % for brown, 63 % for blond, 58 % for black, and 48 % for red hair. From the 5 % of samples (N 0 162) with <90 % call rate, 56 % yielded correct continental ancestry predictions while 7 % yielded sufficient genotypes to allow hair and eye color prediction. Our results demonstrate that the Identitas v1 Forensic Chip holds great promise for a wide range of applications including criminal investigations, missing person investigations, and for national security purposes.
# Introduction
There have been, and likely will continue to be, forensic cases where the evidentiary DNA profile does not directly match that of a known individual or any reference sample profile contained within a national DNA database. In addition, current forensic DNA profiling has provided, and likely will continue to provide, little or no information in a number of missing person cases, including mass disaster identification, where scant information is available on the putative identity of the remains found. Traditional policing places heavy reliance on human eyewitnesses to enable investigators to identify suspects. While eyewitness reports have been shown to be helpful, they are highly error prone [bib_ref] Eyewitness identification: line-ups on trial, Spinney [/bib_ref] [bib_ref] From the lab to the police station. A successful application of eyewitness..., Wells [/bib_ref] , and consequently a number of people convicted on the basis of eyewitness identification evidence have been exonerated through forensic DNA testing [bib_ref] From the lab to the police station. A successful application of eyewitness..., Wells [/bib_ref]. The emerging field of DNA intelligence allows novel investigative leads to be developed directly from the DNA of a forensic sample that can help in identifying persons previously unknown to the authorities. This has the benefit of reducing reliance on human eyewitness accounts and can provide leads in the many cases without known human eyewitnesses [bib_ref] DNA-based prediction of human externally visible characteristics in forensics: motivations, scientific challenges,..., Kayser [/bib_ref] [bib_ref] Improving human forensics through advances in genetics, genomics and molecular biology, Kayser [/bib_ref]. Valuable information in this respect includes biogeographic ancestry and externally visible characteristics (EVC) of the unknown sample donor such as sex, eye color and hair color, and others, via the discipline of Forensic DNA Phenotyping (FDP), as well as the relatedness of the unknown donor with alleged family members.
Worldwide scientific initiatives such as the Human Genome Project [bib_ref] The sequence of the human genome, Lander [/bib_ref] and subsequently the International Hap-Map Project [bib_ref] A map of human genome sequence variation containing 1.42 million single nucleotide..., Sachidanandam [/bib_ref] [bib_ref] The International HapMap Project, Consortium [/bib_ref] [bib_ref] A second generation human haplotype map of over 3.1 million SNPs, Consortium [/bib_ref] [bib_ref] Integrating common and rare genetic variation in diverse human populations, Consortium [/bib_ref] together have laid the foundations for the discovery and large-scale population diversity catalogues of several millions of single nucleotide polymorphisms (SNPs). Highly effective massively parallel SNP genotyping platforms using microarray technology were developed from these resources allowing genome-wide analysis of currently over a million SNPs in a single test [bib_ref] A genome-wide scalable SNP genotyping assay using microarray technology, Gunderson [/bib_ref] [bib_ref] Single nucleotide polymorphism arrays: a decade of biological, computational and technological advances, Laframboise [/bib_ref]. High-resolution SNP microarrays have been used in various human population studies, such as in the worldwide Human Genome Diversity Panel (HGDP-CEPH) [bib_ref] Worldwide human relationships inferred from genome-wide patterns of variation, Li [/bib_ref] [bib_ref] Genotype, haplotype and copy-number variation in worldwide human populations, Jakobsson [/bib_ref] , and in population studies within continents such as Europe [bib_ref] Correlation between genetic and geographic structure in Europe, Lao [/bib_ref] , Asia [bib_ref] Mapping human genetic diversity in Asia, Consortium [/bib_ref] , Africa, India [bib_ref] Reconstructing Indian population history, Reich [/bib_ref] , and Oceania [bib_ref] Demographic history of Oceania inferred from genome-wide data, Wollstein [/bib_ref]. Together with the HapMap data resources [bib_ref] Integrating common and rare genetic variation in diverse human populations, Consortium [/bib_ref] and candidate marker studies including those on Y-chromosomal and mitochondrial DNA diversity [bib_ref] Use of Y chromosome and mitochondrial DNA population structure in tracing human..., Underhill [/bib_ref] , substantial knowledge on DNA-based inference of biogeographic ancestry has begun to be realized. From these datasets, so-called ancestry-informative DNA markers (AIMs) have been developed. Autosomal AIM sets usually provide ancestry resolution at the level of broad geographic regions such as continents [bib_ref] Analyses of a set of 128 ancestry informative single-nucleotide polymorphisms in a..., Kidd [/bib_ref] [bib_ref] Ancestry informative marker sets for determining continental origin and admixture proportions in..., Kosoy [/bib_ref] [bib_ref] A panel of ancestry informative markers for estimating individual biogeographical ancestry and..., Halder [/bib_ref] [bib_ref] Developing a set of ancestrysensitive DNA markers reflecting continental origins of humans, Kersbergen [/bib_ref] [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , whereas with some particular Ychromosomal and mitochondrial AIMs, within-continental resolution can be achieved [bib_ref] Improving human forensics through advances in genetics, genomics and molecular biology, Kayser [/bib_ref] [bib_ref] Use of Y chromosome and mitochondrial DNA population structure in tracing human..., Underhill [/bib_ref]. In some efforts, multiplex AIM panels suitable for forensic applications have been developed for autosomal, Y-chromosomal, and mitochondrial SNPs [bib_ref] Inferring ancestral origin using a single multiplex assay of ancestry-informative marker SNPs, Phillips [/bib_ref] [bib_ref] Evaluating self-declared ancestry of U.S. Americans with autosomal, Y-chromosomal and mitochondrial DNA, Lao [/bib_ref] [bib_ref] Revision of the SNPforID 34-plex forensic ancestry test: assay enhancements, standard reference..., Fondevila [/bib_ref] [bib_ref] MtDNA SNP multiplexes for efficient inference of matrilineal genetic ancestry within Oceania, Ballantyne [/bib_ref] [bib_ref] Multiplex genotyping system for efficient inference of matrilineal genetic ancestry with continental..., Van Oven [/bib_ref] [bib_ref] An efficient multiplex genotyping approach for detecting the major worldwide human Y-chromosome..., Van Oven [/bib_ref] [bib_ref] A multiplex SNP assay for the dissection of human Y-chromosome haplogroup O..., Van Oven [/bib_ref]. However, especially when it comes to recombining autosomal markers, such reduced AIM panels tend to provide less ancestry resolution than high-resolution SNP microarrays [bib_ref] Improving human forensics through advances in genetics, genomics and molecular biology, Kayser [/bib_ref].
Moreover, high-resolution SNP microarrays have been used in genome-wide association studies (GWASs) to discover SNPs involved in human EVCs, most notably eye and hair color [bib_ref] Genome-wide analysis indicates more Asian than Melanesian ancestry of Polynesians, Kayser [/bib_ref] [bib_ref] Genetic determinants of hair, eye and skin pigmentation in Europeans, Sulem [/bib_ref] [bib_ref] A genome-wide association study identifies novel alleles associated with hair color and..., Han [/bib_ref] [bib_ref] Digital quantification of human eye color highlights genetic association of three new..., Liu [/bib_ref]. From these studies, and from candidate gene studies [bib_ref] Blue eye color in humans may be caused by a perfectly associated..., Eiberg [/bib_ref] [bib_ref] A single SNP in an evolutionary conserved region within intron 86 of..., Sturm [/bib_ref] [bib_ref] Sequences associated with human iris pigmentation, Frudakis [/bib_ref] [bib_ref] A polymorphism in the agouti signaling protein gene is associated with human..., Kanetsky [/bib_ref] [bib_ref] Red hair is the null phenotype of MC1R, Beaumont [/bib_ref] , DNA markers predictive for human eye and hair color categories have been identified [bib_ref] Digital quantification of human eye color highlights genetic association of three new..., Liu [/bib_ref] [bib_ref] Eye color and the prediction of complex phenotypes from genotypes, Liu [/bib_ref] [bib_ref] Predicting phenotype from genotype: normal pigmentation, Valenzuela [/bib_ref] [bib_ref] Prediction of eye and skin color in diverse populations using seven SNPs, Spichenok [/bib_ref] [bib_ref] Human eye colour and HERC2, OCA2 and MATP, Mengel-From [/bib_ref] [bib_ref] Gene-gene interactions contribute to eye colour variation in humans, Pospiech [/bib_ref] [bib_ref] Model-based prediction of human hair color using DNA variants, Branicki [/bib_ref]. The first multiplex tools for DNAbased eye and/or hair color prediction have been made available recently [bib_ref] IrisPlex: a sensitive DNA tool for accurate prediction of blue and brown..., Walsh [/bib_ref] [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref] [bib_ref] Verification of eye and skin color predictors in various populations, Pneuman [/bib_ref] [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref] [bib_ref] Further development of forensic eye color predictive tests, Ruiz [/bib_ref] of which at least one, the IrisPlex system for blue and brown eye color prediction [bib_ref] IrisPlex: a sensitive DNA tool for accurate prediction of blue and brown..., Walsh [/bib_ref] , has already been successfully validated for forensic applications [bib_ref] Developmental validation of the IrisPlex system: determination of blue and brown iris..., Walsh [/bib_ref].
All conventional efforts to develop diagnostic tools for DNA intelligence however were restricted to specific elements and were analyzed separately. This limited approach is not only partly due to the incremental nature of the research but also because of technological restrictions on the number of SNPs that could be genotyped reliably in single multiplex assays, suitable for forensic samples. For instance, the SNaPshot chemistry, which is the most widespread SNP typing technology in the forensic genetic field, only allows for the analysis of up to a few dozen SNPs in a single multiplex assay. Thus, several separate DNA tests needed to be performed to enable screening of sufficient markers for comprehensive DNA intelligence efforts. In many forensic cases, however, input DNA amounts are substantially limited, restricting the number of independent DNA tests that can be performed. A single multiplex SNP typing tool including a large number of SNPs is therefore needed that allows various elements of DNA intelligence to be inferred, in parallel, from a single forensic DNA aliquot.
In an international, industry-academic collaboration, the International Visible Trait Genetic (VisiGen) Consortium, the Identitas Version 1 (v1) Forensic Chip was developed. The chip, based on well-established Illumina Infinium technology, allows simultaneous genotyping of 192,658 autosomal SNPs of genome-wide distribution, 3,012 Ychromosomal, 5,075 X-/XY-chromosomal, and 428 mitochondrial SNPs. The genome-wide SNPs were selected primarily for kinship and biogeographic ancestry inference. The panel though was enriched with SNPs that were previously established to have predictive value for biogeographic ancestry and several appearance traits most notably eye and hair color. Herein, the first performance study of the Identitas v1 Forensic Chip is reported based both on data established by consortium members and data from governmental forensic labs in the USA and Canada. A total of 3,196 DNA samples collected from around the world were analyzed. Many of them have recorded sex, continental ancestry, and eye and hair color information, and, in some cases, details of relatedness. The DNA samples were of varying quality and quantity as a result of titration and degradation experiments, and the establishment of mock case-work samples. Genotype quality was assessed, and predictions of sex, biogeographic ancestry, hair color, eye color, and kinship were derived and compared with study-recorded trait data, where available. This study provides the first insights into the performance and feasibility of the Identitas v1 Forensic Chip, the first all-in-one diagnostic tool dedicated to DNA intelligence.
# Material and methods
## Dna samples and available individual information
A total of 3,196 DNA samples were studied. In part, samples were deliberately drawn from a highly biogeographically diverse set of individuals, in order to investigate the quality of prediction of biogeographic ancestry. For a subset of 2,780 individuals, self-reported or site-reported ancestry information was available. For the purposes of obtaining accuracy estimates, individuals were categorized into five major biogeographic groups, plus another category that included, for example, West Asians and individuals from Oceania, for whom HapMap v3 reference samples were not available. The breakdown of site-reported ancestry, where available, and without masking the inevitable overlap in categorizations, was as follows (count in parentheses)-1,880 individuals were categorized as European descent including the following self-declared or site-reported groups: Adygei (25), Austrian (2), Azerbaijani [bib_ref] Blue eye color in humans may be caused by a perfectly associated..., Eiberg [/bib_ref] , British (682), Caucasian [bib_ref] Human eye colour and HERC2, OCA2 and MATP, Mengel-From [/bib_ref] , Chuvash [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , European (380), Georgian (117), German (2), Hungarian [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Irish (303), Italian (2), Komi [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Poland (4), and White (204) and 240 individuals were categorized as East Asian descent including the following self-declared or site-reported groups: Ami [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Asian [bib_ref] Use of Y chromosome and mitochondrial DNA population structure in tracing human..., Underhill [/bib_ref] , Atayal [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Cambodian [bib_ref] Use of Y chromosome and mitochondrial DNA population structure in tracing human..., Underhill [/bib_ref] , Chinese (1), Hakka (25), Japanese [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Korean [bib_ref] Developing a set of ancestrysensitive DNA markers reflecting continental origins of humans, Kersbergen [/bib_ref] , Laotian [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Micronesian [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , and Yakut [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref]. Although Micronesia is in the Western Pacific, Micronesians genetically are known to be largely of East Asian ancestry as result of their migration history (although a Near Oceanian ancestry component exists as well), which justifies their grouping here; 176 individuals were categorized as African descent including the following self-declared or site-reported groups: African (2), African American (3), Barbadian (2), Black (108), Ethiopian [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Hausa [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , and Jamaican [bib_ref] A genome-wide scalable SNP genotyping assay using microarray technology, Gunderson [/bib_ref] ; 123 individuals were categorized as South American descent including the following self-declared or site-reported groups: Karitiana [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Mayan [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Pima [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Quechua [bib_ref] A panel of ancestry informative markers for estimating individual biogeographical ancestry and..., Halder [/bib_ref] , and Ticuna (25); 31 individuals were categorized as South Asian descent including the following self-declared or site-reported groups: Kerala [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] and South Asian (6); and 330 were categorized outside of these major biogeographic groups: Druze [bib_ref] MtDNA SNP multiplexes for efficient inference of matrilineal genetic ancestry within Oceania, Ballantyne [/bib_ref] , Gimi (15), Inuit [bib_ref] Demographic history of Oceania inferred from genome-wide data, Wollstein [/bib_ref] , Iraq (2), Kazakhstani (57), Khanty [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref] , Lebanese (8), Mixed Race (34), Nasioi, Other [bib_ref] Analyses of a set of 128 ancestry informative single-nucleotide polymorphisms in a..., Kidd [/bib_ref] , Uzbek (78), and Yemeni [bib_ref] Proportioning whole-genome single-nucleotide-polymorphism diversity for the identification of geographic population structure and..., Lao [/bib_ref].
The majority of the aforementioned DNA samples came from TwinsUK and the QIMR Twin Registry studies, as well as from the Yale collection as described in detail elsewhere [bib_ref] The UK Adult Twin Registry (TwinsUK), Spector [/bib_ref] [bib_ref] A genome-wide scan for naevus count: linkage to CDKN2A and to other..., Zhu [/bib_ref] [bib_ref] Significant variation in haplotype block structure but conservation in tagSNP patterns among..., Gu [/bib_ref]. For those samples, DNA was derived from whole blood either directly or from bloodderived lymphoblastoid cell lines. Additionally, a subset of 171 DNA samples was derived from more forensically relevant sources which included (counts in parentheses): hair (2), buccal swab (97), blood swab (9), semen (2), vaginal swab (3), saliva (3), mucus (1), gum (3), drink container swab [bib_ref] The International HapMap Project, Consortium [/bib_ref] , cigarette butt (10), chap-stick swab (1), swab of tape ends (1), saliva/blood mixture (1), and vaginal swab/semen mixtures [bib_ref] Multiplex genotyping system for efficient inference of matrilineal genetic ancestry with continental..., Van Oven [/bib_ref]. Two groups contributed sexual-assault type samples. Twenty-nine samples, consisting of either vaginal or buccal samples from female donors (N 0 8), were spiked with varying amounts of semen from male donors (N 0 3) and subjected to a standard differential extraction procedure. Results derived using Plexor HY (Promega), a dual autosomal and male-specific quantification assay, indicated that the percentage of male DNA ranged from 100 % to none. One group submitted an additional mixture made up of vaginal swab DNA plus semen. Sensitivity samples were contributed by two groups. From the first group, four samples were run in a dilution series of total DNA input at 200, 100, 50, 10, 3.3, and 1 ng, leading to a total of 24 samples. The DNA concentration per sample was measured twice using the Quantifiler Human DNA quantification kit (Applied Biosystems) to determine total concentrations of 40 ng/μl down to 200 pg/μl, using a 5-μl sample volume. The second group applied serial dilution to five reference sample extracts of 500 ng to 50 pg total DNA, yielding total DNA concentrations for each sample of 25, 2.5, 0.25, 0.025, and 0.0025 ng/μl. Measurements were taken using Plexor HY (Promega).
Degraded samples were experimentally derived from four pre-extracted DNA samples using (a) titrated DNase treatments and (b) by subjecting samples to ultraviolet (UV) light time courses. Quantified dilutions at concentrations of 20 and 2 ng/μl, in a total sample volume of 110 μl (Quantifiler Human DNA quantification kit, Life Technologies), underwent DNase (Sigma) treatment (0.1 U on each 110-μl volume) for 0, 1, 5, and 10 min with approximately 10 μl of each sample extracted. Thus, three time points and two different concentrations were assessed, for each of the four DNA samples (24 samples altogether). Samples were exposed to UV light for time intervals of 0, 5, 10, and 30 min, using the Bio-Link (Vilber Lourmat) at a strength of 50 J/cm 2 . Three time points for two different concentrations (100 and 10 ng total DNA per 5 μl) were derived for each of four samples (24 samples altogether).
Prior to chip genotyping, DNA samples collated from the different collaborator sites were re-quantified using a Pico-Green-based assay (Life Technologies). Genotyping followed the standard Illumina Infinium iSelect protocol (www.illumina.com). Fluorescence intensities were detected by the Illumina iScan and analyzed using Illumina's Bead-Studio software. The reaction volumes were 2 μl for quality checking and 5 μl for genotyping; additional volume was required to allow for pipetting. In the data received from Illumina, samples were categorized as either "passed" or "failed" using the standard control metrics from Illumina, with failure assigned to samples with more than 10 % missing genotypes calls.
## Statistical analyses
Female-derived DNA was inferred on the basis of Xchromosome heterozygosity; the presence of Y-chromosome genotypes confirmed the presence of male-derived DNA. Biparental biogeographic ancestry predictions were conducted using a marker subset of 81,031 autosomal SNPs exhibiting low correlation (low linkage disequilibrium). Patrilineal ancestry was derived from a subset of 484 Ychromosomal SNPs, and matrilineal ancestry was derived using a subset of 280 mitochondrial SNPs, for which phylogenetic and geographic origin information was available [bib_ref] Updated comprehensive phylogenetic tree of global human mitochondrial DNA variation, Van Oven [/bib_ref] [bib_ref] Y chromosome diversity, human expansion, drift, and cultural evolution, Chiaroni [/bib_ref]. Principal components analysis (PCA) was conducted using, as a reference, data from HapMap version 3 [bib_ref] Integrating common and rare genetic variation in diverse human populations, Consortium [/bib_ref]. Individuals were assigned to continental groups using a simple distance-based clustering algorithm. Model-based clustering, assuming five populations with distinct allele frequencies, was conducted using the same reference set [bib_ref] Inference of population structure using multilocus genotype data, Pritchard [/bib_ref] [bib_ref] Inferring weak population structure with the assistance of sample group information, Hubisz [/bib_ref]. Hair color and eye color were predicted using multinomial logistic regression models of predictive SNPs as described elsewhere [bib_ref] IrisPlex: a sensitive DNA tool for accurate prediction of blue and brown..., Walsh [/bib_ref] [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref] with the difference that the following four hair color-predicting DNA variants from the MC1R gene could not be implemented on the chip: N29insA (INDEL), Y152OCH, rs1805007, and rs1805009; the hair color prediction model used was adjusted accordingly. Degrees of relatedness were inferred for each pair by calculation of the proportion of the genome shared identical by state (IBS) based upon 192,576 autosomal markers with minor allele frequencies greater than 1 %, and less than 5 % missing genotypes [bib_ref] PLINK: a tool set for whole-genome association and population-based linkage analyses, Purcell [/bib_ref].
# Results
## Technical chip performance with relevance for phenotype inference
## Genotype reproducibility
Concordance testing using different SNP microarray platforms was performed on 102 QIMR samples genotyped in this study on the Identitas v1 Forensic Chip and previously, at a different laboratory, on the Infinium 610-Q (Illumina) GWAS arrays [bib_ref] Common variants in the trichohyalin gene are associated with straight hair in..., Medland [/bib_ref]. Up to 107,262 SNPs directly overlap between both arrays. The observed discordance was 92 out of the total of 10,831,289 genotype calls (rate 0.00085 %, data available on request). Although, these data do not show which of the two microarrays produced the bona fide genotype, the high concordance rate of >99.999 % indicates that the reproducibility of genotypes from the Identitas v1 Forensic Chip, at least for samples of similar quality and quantity as the 102 tested here, is very high.
## Sensitivity testing
Serial dilutions were conducted by one group for five DNA samples extracted from buccal samples of five individuals of European biogeographic origin. Each was diluted to contain 175, 17.5, 1.75, 0.175, or 0.0175 ng in the 7-μl reaction volume for chip genotyping. For the lowest concentration of 0.0175 ng reaction DNA, all five samples failed platform QC with overall genotype call rate <90 %, none had the full complement of markers for hair and eye color prediction, and only one provided an accurate prediction of biogeographic ancestry (quantitative method, see later). At the next level of DNA concentration, 0.175 ng reaction DNA, one sample passed platform QC, one had the full complement of markers for hair and eye color prediction, and three provided accurate predictions of biogeographic ancestry. At the next level, 1.75 ng reaction DNA, three passed platform QC, three had the full complement of markers for hair and eye color prediction, and all five provided accurate predictions of biogeographic ancestry. At higher concentrations, amounting to 17.5 and 175 ng total DNA, 9/10 passed platform QC, 8/10 had the full complement of markers for hair and eye color prediction, and all 10 provided accurate predictions of biogeographic ancestry. Serial dilutions were conducted by a second group for four DNA samples extracted from blood of four individuals of European biogeographic origin. Each was diluted to 200, 100, 50, 10, 3.3, and 1 ng in a 5-μl volume. It is noted that for this set, further dilution was required to increase volume prior to genotyping, so the amount of DNA used may be lower. For samples at the lowest concentration (1 ng reaction DNA), all failed platform QC, none had the full complement of markers for hair and eye color prediction, but all four provided an accurate prediction of biogeographic ancestry. At the next level of 3.3 ng reaction DNA, one passed QC, one had the full complement of markers for hair and eye color prediction, and all four provided accurate predictions of biogeographic ancestry. At the higher concentrations of 10, 50, and 100 ng total DNA, all samples passed QC, all had the full complement of markers for hair and eye color, and all provided accurate predictions of biogeographic ancestry.
Despite the preliminary character of the sensitivity testing performed here with small sample sizes, these results demonstrate that biogeographic ancestry may be accurately predicted from as little as 1.75 ng DNA or even, in some cases, as little as 0.175 ng DNA. For other traits, predictive success is dependent on generating sufficient relevant genotypes.
## Degradation testing
Twenty-four samples derived from four initial DNA samples were subjected to severe ultraviolet degradation. Upon genotyping, only three passed platform quality checks. They corresponded to three of the 100 ng samples at the first time point, and all had accurate predictions of biogeographic ancestry and sufficient genotypes to allow hair and eye color prediction. The remaining 21 samples had between 18 and 50 % missing genotypes, none had sufficient genotypes to allow hair and eye color prediction, but four (all 10 ng at first time point) provided accurate predictions of biogeographic ancestry.
The same 24 samples were subjected to less severe, enzymatic degradation. Median concentration for the set was <1 ng/μl. Upon genotyping, five samples failed platform quality checks: one at a degradation time of 1 min, two at 5 min, and three at the final time point of 10 min. The five samples had between 11 and 21 % missing genotypes, compared to less than 10 % missing genotypes in the 19 samples which passed platform quality checks. The elevated failure rate in this small degradation subset confirms the damaging effect of nucleases on DNA genotyping performance. A total of 20/24 samples however had sufficient genotypes to allow prediction of hair color and eye color, and all 24 enzymatically degraded samples led to accurate estimates of biogeographic ancestry.
## Analysis of forensic-type samples
A total of 141 single-source DNA samples extracted from hair, buccal swab, blood swab, semen, vaginal swab, saliva, mucus, gum, drink container swab, cigarette butt, chap-stick swab, and swab of tape ends were examined, to assess performance in more typical case-work samples. DNA concentration was measured by PicoGreen and found to be generally low (median 0 2.7 ng/μl). Of these 141 samples, 102 passed QC and had median PicoGreen-based concentration of 3.1 ng/μl (range 0.7-56.6 ng/μl). A total of 102/ 141 (72 %) samples had sufficient genotypes to allow prediction of hair color and eye color. Biogeographic ancestry was available for 125 of the samples, the majority (85 %) of whom were of European ancestry. A total of 110 out of 125 (88 %) of the predictions were correct on the level of continental ancestry. There were no clear differences in performance among sample types; correct predictions were obtained from blood swab, buccal swab, semen, vaginal swab, mucus, gum, drink container swab, and hair.
In addition, a total of 30 multisource DNA samples from simulated sexual assault material extracted after differential lysis were examined. Again, DNA concentration tended to be low (median 0 1.6 ng/μl, range 0.8-61.4 ng/μl). A total of 19/30 samples passed quality checks; however, upon unblinding at source, it emerged that only 13 of them contained male DNA. For all 13 samples, the Y-chromosome haplogroup was obtained and their assumed geographic region of origin was found to be consistent with site-reported biogeographic ancestry.
Chip-based inference of sex, ancestry, appearance, and relatedness Across the whole study, a total of 3,034 (95 %) samples passed platform quality checks with overall genotype call rates >90 %, while 162 samples (5 %) failed this threshold. In the following sections, results are presented of the analysis of the 3,034 DNA samples that passed quality control checks.
## Inference of sex
A two-pronged approach was taken for chip-based prediction of whether the DNA used was derived from a man or a woman. Y-chromosome haplogroups, derived from known non-recombining male-specific SNPs, were obtained for 1,114 DNA samples, indicating that they were derived from males. In addition, X-chromosome heterozygosity was determined for all samples on the basis of 5,066 X-chromosomespecific markers (without a homologue on the Y chromosome), with an estimated mean heterozygosity of <0.2 implying male-derived DNA and an estimated mean heterozygosity of >0.8 implying female-derived DNA [bib_ref] PLINK: a tool set for whole-genome association and population-based linkage analyses, Purcell [/bib_ref]. Intermediate values were considered inconclusive. Twelve conflicts, between chip-predicted and site-reported sex information, were obtained in the 1,588 samples with site-reported sex information available. Four samples were predicted to be derived from males on the basis of both identified Y-haplogroup and low Xchromosome heterozygosity but were unblinded as being derived from females from the records. Seven samples were predicted to be derived from females on the basis of no inferable Y-haplogroup and high X-chromosome heterozygosity but were unblinded as being derived from males from records. Further nongenetic sex data could not be obtained for these 11 individuals. A plausible explanation however is that DNA mix-ups at some stage may have occurred for these samples, which would correspond to a male-female sample mix-up rate of 0.69 % in our study. One of the 1,588 samples, unblinded as female, was wrongly predicted to be male on the basis of low X-chromosome heterozygosity alone but had no Y-chromosome haplogroup inferable. This sample would have represented a sex misclassification based solely on the X-chromosome heterozygosity approach, emphasizing the importance of using data from both X-and Y-chromosome SNPs for inferring sex from these chip data.
## Inference of continental biogeographic ancestry
A three-dimensional PCA plot of the HapMap 3 reference data [bib_ref] Integrating common and rare genetic variation in diverse human populations, Consortium [/bib_ref] led to the clustering of individual samples into five, somewhat separated, main descent groups: European descent (CEU, TSI), African descent (ASW, LWK, MKK, YRI), East Asian descent (CHB, CHD, JPT), South Asian descent (GIH), and South American descent (MEX) [fig_ref] Figure 1: Three-dimensional principal component analysis plot for a DNA sample from a European... [/fig_ref]. This reference dataset was then used to classify the study samples according to their biparental continental biogeographic ancestry via a distance-based clustering algorithm. For example, the black cross in [fig_ref] Figure 1: Three-dimensional principal component analysis plot for a DNA sample from a European... [/fig_ref] represents an individual sample, unblinded as "White" classified by self-reported ancestry information, which appears very close the European-descent HapMap reference samples (CEU and TSI), highlighting the most likely European biparental genetic ancestry of this individual. Selfor site-reported biogeographic ancestry on the continental level was available for 2,688 out of 3,034 samples. Overall, PCA clustering led to good prediction accuracy for certain categories of biparental genetic ancestry. Predictions of European ancestry were 97 % consistent with self-declared or sitereported ancestry, predictions of African ancestry were 88 % consistent with self/site report, and predictions of East Asian ancestry were 97 % consistent with self/site report. Predictions of South Asian ancestry, however, were lower (69 % consistent with self/site report), while predictions of South American ancestry were poor. Specifically 39 % (52/134) of the predictions of South American ancestry were self-/site-categorized outside of the main five major biogeographic ancestry groups (they were 19 Khanty, 9 mixed race, 9 Other, 1 Lebanese, 1 Kazakhstan, 5 Uzbekistan, and 8 Inuit Aborigine).
A quantitative approach was therefore pursued, which led to the assignment of a probability for each of the five reference groups, allowing more accurate inferences to be drawn. By this quantitative approach, samples were assigned to a single continental ancestry group whenever the probability for that group was greater than 0.70. When the maximum probability for any single continental group was ≤0.70, the sample was assigned to "multiple groups." In this way, 89 % of samples were assigned to a single continental or subcontinental group; the remainder were assigned to multiple defined groups. Details of the accuracy of these predictions are given in the following paragraphs.
Out of a total of 1,877 predictions of European ancestry, 93 % were correct as given by self-report/site report of Adygei, Austrian, Azerbaijani, British, Caucasian, Chuvash, European, German, mixed German/Polish, Georgian, Hungarian, Irish, Komi, Polish, or White (meaning likely European). Of the 128 individuals whose predicted European ancestry was not clearly consistent with self-report/site report, 26 were Druze, 24 were Yemenite, 15 were mixed Italian/ Greek/ Syrian, 15 were Uzbeks, 13 were Kazakhstani, 11 were Inuit, 8 were Lebanese, 4 were mixed race, 4 were "Other," 3 site-reported as African-American, 2 were Iraqi, 1 was Japanese, 1 was Quechua, and 1 was South Asian. Y-chromosome haplogroups were obtained for 52 out of these 128 samples that showed inconsistency between ancestry group inferred from the genome-wide SNPs data and self-/site-recorded ancestry information. Y haplogroups for 49 of these samples are found in people of European paternal origin. Furthermore, the majority of these 128 samples [fig_ref] Figure 1: Three-dimensional principal component analysis plot for a DNA sample from a European... [/fig_ref] , including three site-reported African-Americans, had mitochondrial haplogroups which indicated Western Eurasian maternal origin. These Y and mtDNA results, together with the genome-wide results, suggested that despite self-/sitereported ancestry information, many of these 128 individuals share a considerable proportion of their genetic ancestry with Europeans, likely due to recent European admixture. Some may represent cases were continental ancestry is difficult to infer from genetic data because the individuals descended from a geographic area located between major geographic regions (such as from the Middle East or western parts of Asia).
Out of a total of 233 predictions of East Asian ancestry, 94 % were correct as given by self-/site-reported ancestry of Ami, Asian, Atayal, Cambodian, Chinese, Hakka, Japanese, Korean, Laotian, Micronesian, or Yakut. Of the 14 individuals whose predicted East Asian ancestry was not clearly consistent with self-report/site report, one was site-reported as British, two were Gimi, one was Inuit, one was mixed (Italian/Greek/Syrian), one was Kazakh, one was Uzbek, three were Nasioi, and four were Other. The mitochondrial haplogroups for these individuals, except the Gimi and the Nasioi, indicated Eastern Eurasian maternal ancestry suggesting, together with the genome-wide results, considerable East Asian admixture. Six Y-chromosome haplogroups were obtained and all indicated East Asian, Southeast Asian, or Oceanic origin. The finding that mitochondrial and Y-chromosome haplogroups correctly assign Oceanic origin in the Gimi and Nasioi may illustrate the limitations in the resolution level of the genome-wide data in terms of separating Oceanians from East Asians using our approach. This is likely because Oceanic samples are missing in the HapMap 3 reference data used as reference for this analysis.
Of the total of 145 predictions of African ancestry, 88 % were correct as given by self-/site-reported ancestry of African, African-American, Barbadian, Ethiopian, Hausa, or Jamaican. Of the 17 individuals whose predicted African ancestry was not clearly consistent with self-report/site report, 15 were site-reported as "British," 1 site-reported as Other, and 1 site-reported as White. Without exception, all these 17 samples carried African mitochondrial haplogroups indicating, together with the genome-wide results, considerable African admixture in these samples. Furthermore, all of the four males had African Y haplogroups.
Of the total of 107 predictions of South American ancestry, 98 % were correct as given by self-/site-reported ancestry of Karitiana, Mayan, Pima, Quechua, or Ticuna. Both of the individuals whose genome-wide predicted South American ancestry was not clearly consistent with self-/site-reported ancestry and who classified themselves as Other had mitochondrial haplogroups of Eastern Eurasian origins. Both were female, so Y-chromosome haplogroups were not available.
Of the 24 predictions of South Asian ancestry, 96 % were correct as given by self-/site-reported ancestry of Keralite, mixed Indian/Pakistani, or South Asian. The individual, whose genome-wide predicted South Asian ancestry was not consistent with self-report/site report, was site-reported as British and carried a Western Eurasian mitochondrial haplogroup. This individual was female, so a Y-chromosome haplogroup was not available.
Our approach was insightful also in terms of genetically classifying individuals of mixed continental ancestry. Unfortunately, only a handful of such individuals were unblinded with details of their parental origin. Taking a single example, shows the quantitative assessment of an individual whose father was of African descent and whose mother was of European descent according to the record information. The almost equal proportions of African and European DNA were accurately captured by the method. Furthermore, the determination of the Y-chromosome haplogroup as E-U290, which is observed in Africa, Western Asia, and Europe, and the mitochondrial haplogroup HV, which is observed in Western Eurasia, indicated that the paternal line is African while the maternal line is European, in agreement with record-based ancestry information.
Further useful insights were obtained. [fig_ref] Figure 3: Box plot for quantitative assessments of biogeographic ancestry for 24 Ethiopian individuals... [/fig_ref] shows the box plot for 24 individuals of Ethiopian descent. There is a substantial European component compared with the mainly West African reference samples, indicating the genetically admixed situation of North Africans. Similarly, shows the box plot for 57 individuals from Kazakhstan, which lies on the silk route between China and Europe. It can be seen that individuals who come from a location that is intermediate between the origins of the reference groups may be indistinguishable from an individual of mixed-race origin, using these first-pass techniques. The ongoing development of methods to estimate, for example, LD block size in individuals of mixed ancestry may elucidate the matter further. shows the breakdown of chip-predicted versus sitereported eye color for samples which passed platform quality checks and had both site-reported eye color, as well as a complete genotype profile for the six SNPs required [fig_ref] Figure 1: Three-dimensional principal component analysis plot for a DNA sample from a European... [/fig_ref]. It can be seen that 70 % of predictions of blue eyes and 85 % of predictions of brown eyes agreed with sitereported eye color using the p > 0.7 threshold recommended previously [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref]. [fig_ref] Table 2: Chip-predicted versus site-reported hair color for 1,137 individuals with chip and record... [/fig_ref] shows the breakdown of predicted versus sitereported hair color for samples which passed platform quality checks and had both site-reported hair color as well as the complete genotype profile of the 18 SNPs required (N 0 1,137). It can be seen that using the previously developed prediction guide [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref] , 58 % of predictions of black/dark brown hair corresponded to site report of black, dark brown, or brown hair; 72 % of predictions of brown/light brown/dark Quantitative assessment of biogeographic ancestry from an individual whose father was of African origin and whose mother was of European origin
## Inference of eye and hair color categories
# Inference of relatedness
The proportion of the genome shared IBS was estimated for all pair-wise combinations of the 3,034 samples, using 192,576 markers. For the majority of the samples however, the true relationships were not known/site-reported. The true relatedness of samples was available from one source, for which 3,240 pair-wise comparisons had been performed for 81 samples. In this set, all 27 first-degree relative pairs, ten second-degree relative pairs (four uncles/aunts, five grandparents, and one half-sibling), three third-degree relative pairs (two first cousin pairs and one great aunt), and 3,199 unrelated pairs were correctly identified. One additional pair of individuals was observed to share 9 % of the genome IBS, in line with a fourth-degree relationship (e.g., first cousin once removed). Site report for the two was unrelated, but both were of Caribbean origin. It is acknowledged that the prediction of fourth-degree relationships is only valid in highly outbred populations. For populations which are, or have been in the past, genetically isolated, there will be an overprediction of distant relatives.
# Discussion
The Identitas v1 Forensic Chip is the first all-in-one diagnostic tool targeted for DNA intelligence purposes, allowing for massively parallel genome-wide inference of ancestry, appearance, relatedness, and sex. This DNA chip, manufactured by Illumina using their well-established Infinium technology, is able to deliver highly reproducible genotypes as indicated by the >99.999 % genotyping agreement achieved in an independent comparison with the Infinium 610-Q (Illumina) GWAS array. v1 of the Identitas Forensic Chip exhibited, from samples that passed the quality control threshold of >90 % overall genotype call rate, high predictive power for inference of sex, continental biogeographic ancestry, individual relatedness up to third-degree level, and somewhat less power also for eye and hair color. Even with <90 % call rate, high predictive assignment success was achieved for continental ancestry, although the number of incorrect inferences increases with decreasing overall call rate. The power of prediction of continental biogeographic ancestry may be explained by the large number of array SNPs used for inference, and the partial redundancy of ancestry information across such SNPs. This also was true for the chip-based inference of relatedness, which was based on an even larger number of SNPs, and for the sexinference, based on >5,000 X-chromosomal SNPs. The situation is very different for chip-based eye and hair color prediction, where only a small number of particular nonredundant SNPs with high predictive value were used, i.e., six for eye color prediction and 18 for hair color prediction. As long as these particular SNPs are genotyped correctly, eye and hair color prediction can be achieved. The overall call rate therefore provides only a first indication of the chip's practical performance on ancestry, relatedness, sex, and eye/ Developing a detailed knowledge about the direct and ideal relationship between DNA quantity and quality, chip performance, and the accuracy of ancestry, relationship and eye/hair color inference requires additional tailored datasets to be generated in the future. One technical complication in the current study was the limited availability of PCR-based sample quantification prior to genotyping. For the majority of samples, PicoGreen provided the only measure of DNA concentration available. Although PicoGreen is known to be reliable in measuring higher DNA concentrations, i.e., tens of nanograms per microliter and beyond, it tends to be less accurate in the single nanogram and sub-nanogram per microliter range encountered in many forensic applications. Similarly, the tolerance of the Identitas v1 Forensic Chip in using degraded DNA for successful inference of ancestry, appearance, and relatedness needs to be followed up with more extensive testing. Based on our preliminary data, the chip genotyping of partially degraded DNA still allowed high predictive value for biogeographic ancestry, but more data are needed to develop clear degradation thresholds. As with DNA quantity, the large number of SNPs used for ancestry, relatedness, and (less so) inference of sex worked in favor of the chip when dealing with degraded DNA.
One challenge in conducting our analyses was the consolidation of data from multiple sources, each with different conventions for recording biogeographic ancestry information. The large geographic categories described here are necessarily simplifications and the accuracy estimates are likely to be conservative. For instance, a total of 682 individuals were British by self-report/site report and it was clear, upon examination of haplogroups, that this term was, in some instances, intended in the sense of "nationality" rather than biogeographic origin. Our genome-wide analysis however grouped them with individuals of true European origin which consequently lowered the accuracy rate obtained for European ancestry assignment. Other simplifications included the categorization of Mayans with South Americans, as well as the categorization of Micronesians with East Asians. The motivation in providing such wide groupings was to maximize the use of the available data.
The estimates of prediction accuracy for eye and hair color obtained here were lower than those previously reported with the IrisPlex system for eye color [bib_ref] IrisPlex: a sensitive DNA tool for accurate prediction of blue and brown..., Walsh [/bib_ref] [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref] and the HIrisPlex system for hair color prediction [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref]. Whereas all six IrisPlex SNPs together with the IrisPlex eye color prediction model [bib_ref] IrisPlex: a sensitive DNA tool for accurate prediction of blue and brown..., Walsh [/bib_ref] [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref] were applied here without modification, only 18 of the previously reported 22 HIrisPlex DNA variants for hair color prediction [bib_ref] Model-based prediction of human hair color using DNA variants, Branicki [/bib_ref] [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref] could be implemented in the Identitas v1 Forensic Chip. An adjusted hair color prediction model was therefore used. The four DNA variants missing on the chip lay in the MC1R gene and are known to be predictive of both red hair and dark hair [bib_ref] Model-based prediction of human hair color using DNA variants, Branicki [/bib_ref] [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref]. Indeed, nearly 60 % of individuals with site-reported red hair were missed. This strongly contrasts with only 14 % instances of red hair missed using the HIrisPlex system containing all 22 DNA variants [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref]. The mis-categorization of those individuals with red hair had a knock-on effect on the accuracy of other predicted hair color categories. Secondly, in contrast to the previous eye and hair color prediction studies that solely or mainly used single grader color classifications [bib_ref] Model-based prediction of human hair color using DNA variants, Branicki [/bib_ref] [bib_ref] IrisPlex: a sensitive DNA tool for accurate prediction of blue and brown..., Walsh [/bib_ref] [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref] [bib_ref] The HIrisPlex System for simultaneous prediction of hair and eye colour from..., Walsh [/bib_ref] ], the current study used self-reported eye and hair color. The inevitable subjectivity of self-report means that the estimates achieved here are likely to be conservative. For instance, in two different European studies where single-grader eye color phenotyping was applied, the intermediate eye color category was observed at frequencies of 9.6 % [bib_ref] Eye color and the prediction of complex phenotypes from genotypes, Liu [/bib_ref] and 14 % [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref]. These values are considerably lower than the 35 % of self-reported eye colors that fall into the intermediate category from the present study. Previous studies demonstrated that the intermediate eye color category with the six IrisPlex SNPs used here cannot be predicted with as high accuracy as blue and brown eyes [bib_ref] Eye color and the prediction of complex phenotypes from genotypes, Liu [/bib_ref] [bib_ref] IrisPlex: a sensitive DNA tool for accurate prediction of blue and brown..., Walsh [/bib_ref] [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref]. The inflation of the intermediate eye color category is caused by self-reporting errors and also has an impact on the estimated prediction accuracies for blue and brown. Notably, if we exclude the selfreported intermediate eye color individuals from the prediction analysis, we receive much higher accuracies at 90 % for blue and 94 % for brown eye color. These values are similar to the 94 % accuracy for blue and brown achieved in the previous study using single-grader eye color phenotypes [bib_ref] DNA-based eye colour prediction across Europe with the IrisPlex system, Walsh [/bib_ref]. The greatest value of the Identitas v1 Forensic Chip relative to other tools for FDP or tools for other aspects of DNA intelligence is that the various targets for ancestry, appearance, relatedness, and sex are combined in a single all-in-one diagnostic tool. In criminal investigations, in the absence of a match with a reference sample, such insights can dramatically focus the downstream investigations. The DNA-based investigative intelligence obtained can be used in conjunction with, or in the absence of human eyewitness information, to potentially lead to the identification of suspects. The highly accurate inference of relatedness opens further avenues of application, including paternity/relationship resolution, homeland security matters, and the resolution of missing person investigations, through the analysis of found human remains, including in cases of mass disasters. All usage of this (and similar) technology must, of course, comply with the legal requirements of the country in which it is aimed to be applied to practical forensic case work. The technology introduced here therefore offers a novel opportunity of actionable data in a variety of no-match scenarios. In addition, this comprehensive chip requires the consumption of only one aliquot of DNA evidence material. The proposed tool will be a valuable complement to crime-scene short tandem repeat analysis which represents the industry standard for DNA-based identification through direct matching.
Further developments are underway. The current Version 1 of the Identitas Forensic Chip already contains SNPs associated with freckles, moles, curly hair, skin color, earlobe shape, and body height. However, the phenotype prediction values of the currently known markers for these EVCs are not high enough to be practically useful: more predictive DNA markers need to be identified. Subsequent versions of the Identitas Forensic Chip will include additional markers for these and other appearance traits, as they are identified. For instance, the first GWAS studies on facial shape features have just appeared in the literature [bib_ref] Genome-wide association study of three-dimensional facial morphology identifies a variant in PAX3..., Paternoster [/bib_ref] [bib_ref] A genome-wide association study identifies five loci influencing facial morphology in Europeans, Liu [/bib_ref]. Future improvements may also include refining eye and hair color prediction, especially for the more intermediate colors and eventually moving from the current categorical approach to the prediction of continuous shades of eye and hair color.
Many of the geographically diverse samples that were used in the current study can now serve as reference datasets for any future analysis of biogeographic ancestry using the Identitas v1 Forensic Chip. Although not assessed here, the v1 chip already contains SNPs that carry subregional ancestry information such as for Europe, and downstream efforts will include subregional ancestry into the prediction pipeline. This chip and future iterations will be a valuable addition to the forensic geneticist's toolkit.
Acknowledgments The International VisiGen Consortium wishes to acknowledge the contributions of their research institutions, study investigators, field staff, and the study participants, in creating the scientific resources accessed. We also would like to thank the scientists at Illumina for the help with assay scoring, assay selection, and genotyping. TwinsUK is supported by the Wellcome Trust and the NHS NIHR Biomedical resource grant to Guys and St. Thomas' foundation hospitals and KCL. Identitas Inc. sponsors VisiGen via a research grant to a number of the respective academic institutions.
Conflict of interest BK had a financial relationship with Identitas Inc., ATB, and LR have a financial relationship with Identitas Inc. TDS and MK had consulted for Identitas Inc. and are on the SAB but without financial or other direct benefits. All other authors declare that they have no conflict of interest.
Disclaimer SNP genotyping was supported in part by the FBI Laboratory Division. Names of commercial manufacturers are provided for identification only and inclusion does not imply endorsement of the manufacturer or its products or services by the FBI. The views expressed are those of the authors and do not necessarily reflect the official policy or position of the FBI or the US Government. This manuscript was filed under the number 12-18 at the Counterterrorism and Forensic Science Research Unit, Federal Bureau of Investigation Laboratory Division.
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
[fig] Figure 1: Three-dimensional principal component analysis plot for a DNA sample from a European individual denoted by a black cross, together with reference data from HapMap v3 i.e., from individuals of European descent (CEU Utah residents with Northern and Western European ancestry from the CEPH collection; TSI Tuscans in Italy), African descent (ASW African ancestry in Southwest USA; LWK Luhya in Webuye, Kenya; MKK Maasai in Kinyawa, Kenya; YRI Yoruba in Ibadan, Nigeria), East Asian descent (CHB Han Chinese in Beijing, China; CHD Chinese in Metropolitan Denver, CO; JPT Japanese in Tokyo, Japan), South Asian descent (GIH Gujarati Indians in Houston, TX), and South American descent (MEX Mexican ancestry in Los Angeles, CA) [/fig]
[fig] Figure 3: Box plot for quantitative assessments of biogeographic ancestry for 24 Ethiopian individuals blonde corresponded to site report of dark brown, brown, light brown, or dark blonde hair; 63 % of predictions of blonde/ dark blonde corresponded to site report of light brown, dark blonde, or blonde hair; and 48 % of predictions of red hair corresponded to site report of red hair. [/fig]
[table] Table 2: Chip-predicted versus site-reported hair color for 1,137 individuals with chip and record information available a Using the prediction model and the prediction guide described elsewhere[52] Predicted hair color a Site-reported Hair Color Total [/table]
|
Let's Get Back to Normal? COVID-19 and the Logic of Cure
The COVID-19 pandemic has inversed certainties of absolutes of cure in everyday life but paradoxically this has occurred during a time when novel scientific advancements seem to herald a new frontier of cures for rare diseases, chronic conditions, disabilities and viruses that were previously incurable. In this paper, I illustrate the development of a logic of cure by first of all noting a lacuna in the medical sociological and anthropological literature, where although a lot of empirical research and theoretical work to understand cure has been undertaken, there has been no sociology or anthropology of cure. Using three case studies, I examine what they reveal about the logic of cure. Firstly, I argue that there is a development of a bioethics of cure in reactions of disability community and disabled people to care as cure during the COVID-19 pandemic. The second case-study focuses on understanding limitations of vaccines and how people react against such indeterminancies of loss of absolutes of cure. Lastly, the final case study describes how while there are cures, for example, for rare genetic conditions, they are often initially curated with long-term cost-benefit analysis for the Global North. In conclusion, it is found that many of the developments within sociology and anthropology are missing from a logic of cure and that a new theory of cure has to develop.
# Introduction
The COVID-19 pandemic has brought to the fore the importance of critically examining our responses to accelerated scientific developments of "cures". From the differing socio-cultural reactions in the Global North encompassing broad calls to get vaccinated and back to "normal" life to the hesitancy of risk and historical suspicion of scientific experimentation in some communities. As well as lack of choice of "cures" in inequalities in who gains first access to diagnostics, medicines, therapies and vaccines in the Global North and South, the pandemic has revealed not only the complexities of a curative imperative but also how the general public has coped with a loss of certainty of what "cure" now means.
In popular culture, what it meant to be incurable or have no possibility of recovery or cure, used to be reserved for serious physical or mental illnesses, chronic conditions, rare diseases, disabilities or viruses. The realm of everyday health and "normal" was not part of such medical pathology and environmental risks, thus "cure, " recovery, healing or convalescence from an illness or disease to "health", did not need to be unduly examined. The very idea that there could be different forms of "cure" or of ways in which your body and mind recovered was never part of that analysis. Nor did cures encompass disability, disease or mental illness. Or even it's opposite, in temporal understandings of life where cure was never a given. The emphasis has always been on the "curing" or how people recovered and were cared for, without description of the diversity of experiences of cure itself. Cure was always seen in terms of an end-point, rather than along a continuum of constant illness and possible or impossible recovery cycles, each with differing forms of cure during the life-course.
Despite warnings from the scientific community of future possibilities of anti-microbial resistance and epidemics of Zika, MERS-Cov and Ebola destabilizing our understandings of cure, illness and disability; that there would always be cures and recovery to health in the Global North seemed to be a popular given. While several successful vaccines have been found for COVID-19 and vaccination campaigns begun in many countries, this has initially been mainly in the Global North with localized priorities and vaccine nationalism often triumphing over global rights to health. Reactions to vaccine developments have also been dampened by noting how population engagement is critical, immunity might not be experienced similarly and that new variants of the SARS-CoV-2 virus may develop, which will again need fast vaccine development or vaccine boosters.
The COVID-19 pandemic has thus inversed certainties of absolutes of cure in everyday life but paradoxically this has occurred during a time when novel scientific advancements seem to herald a new frontier of cures for rare diseases, chronic conditions, disabilities and viruses that were previously incurable. For instance, innovative developments in vaccines can now target the structure of immunogen designs to ensure immunity against influenza, gene-based vaccine platforms like mRNA can target respiratory syncytial virus, and using recombinant proteins can ensure control of latent tuberculosis infection [bib_ref] Novel vaccine technologies for the 21st century, Mascola [/bib_ref]. Similarly, the development of the Moderna mRNA vaccine has had a ripple effect in causing research excitement that a successful vaccine for HIV could be developed using the same insights [bib_ref] In the era of mrna vaccines, is there any hope for hiv..., Esteban [/bib_ref]. Some of these current developments build on scientific and technological advances, for example in genomics and genetics, that mean a better understanding of diseases, infections and their prevention. Due to the increase of high-dimensional biology, omics testing (genomics, proteomics and metabolomics), screening services, use of big data analytics, increase of biomonitoring and the move toward personalized medicine based on genes (genomics), mRNA (transcriptomics), proteins (proteomics) and metabolites (metabolomics), there are advances in early identification, prevention as well as innovation of technologies of cure. For example, in pharmacogenomics, direct to consumer genetic testing or the understanding of how inheritance can predict biological responsiveness to particular medicines [bib_ref] Omic'technologies: genomics, transcriptomics, proteomics and metabolomics, Horgan [/bib_ref] [bib_ref] From big data analysis to personalized medicine for all: challenges and opportunities, Alyass [/bib_ref] [bib_ref] Blood transcriptomics and metabolomics for personalized medicine, Li [/bib_ref] [bib_ref] Biomonitoring in the era of the exposome, Dennis [/bib_ref].
In the above, a "logic of cure" that places as central an imperative to cure, which has become normalized and commodified is evident. Just ascontrasted a patient centered "logic of care" with one of consumer "choice" in medicine, I argue that these scientific and technological advances have divested both care and cure to the needs of the marketplace. While the material of the human body seems to be at stake, cures bring into focus how the biological, socio-cultural and environmental become entangled because of their economic and political impacts. Hence, [bib_ref] Situating local biologies: anthropological perspectives on environment/human entanglements, Niewöhner [/bib_ref] argued that with epigenetic advances, there had to be recognition of how the relationship between nurture and nature had changed in situating local biologies. The neoliberal promises of late modernity have been incorporated into such a curative imperative and patients, health care professionals, industry, academic as well as philanthropic investors now act together with politicians and national interests to ensure economic, political and socio-cultural momentum around developments of cures, such as during the COVID-19 pandemic. This reveals stratified inequalities in not only what will be cured but also who, when and where curative processes will take place and why.
In this paper, I illustrate the development of a logic of cure by first of all noting a lacuna in the medical sociological and anthropological literature, where although a lot of empirical research and theoretical work to understand cure has been undertaken, there has been no explicit sociology or even anthropology of cure [bib_ref] Who gets cured? COVID-19 and developing a critical medical sociology and anthropology..., Berghs [/bib_ref]. Using three case studies, I examine what they reveal about the logic of cure. Firstly, I argue that there is a development of a bioethics of cure in reactions of disability community and disabled people to care as cure during the COVID-19 pandemic. The second case-study focuses on limitations of COVID-19 vaccines and how people react against such indeterminacies of cure. Lastly, the final case study describes how while there are genomic cures, for example, for rare genetic conditions, they are often curated with cost-benefit analysis. In conclusion, it is discussed what the three case studies reveal about the development of a sociology and anthropology of cure.
## Why no sociology of cure?
If you examine the plethora of sociological and corresponding philosophical and ethical literature on care, there has been work on "cure" but it seems odd that cure has not been given the same explicit attention as care? For instance, there is no philosophy of cure but a philosophy of care is a well-grounded discipline. Despite the proliferation of the term cure and its link to care in clinical terms, not much has been explicitly written in sociological research theoretically investigating cure. What an accelerated curative imperative now entails? How that feels as a patient or for a family if you have had a life-long or life-threatening illness and been "functionally" cured? What technologies and commercial investments become implicated, in what biological materials, and if and how the future changes if your "curative hopes" have been realized or ruined? Generally, there is a lacuna about "cure" but that is correlated to how we understand cure as becoming ever more specialized in particular therapies, medicines and interventions. Cure is also scientifically developed as innovative "hype" to be critically sociologically appraised, in keeping with new understandings of self, body and environment developed from the 1980s onwards.
## Understanding ourselves?
Sociological lay understandings and experiences of health and illness from the 1980s onwards have typically focused on narratives of loss when faced by illness, disease and disability.
For instance, in "biographical disruption" [bib_ref] Chronic illness as biographical disruption, Bury [/bib_ref] , "loss of self " [bib_ref] Loss of self: a fundamental form of suffering in the chronically ill, Charmaz [/bib_ref] to "narrative reconstruction" [bib_ref] The genesis of chronic illness: narrative reconstruction, Williams [/bib_ref] , and the ways in which lay understandings of the "self " are confronted with biomedical explanations of illness, disease and disability when there is no cure [bib_ref] Lay experiences of health and illness: past research and future agendas, Lawton [/bib_ref]. Furthermore, conceptions of the body, selves and embodiment underwent further rapid changes from medicalization, (bio) medicalization to genetization with the start of the 1999 Human Genome Project [bib_ref] Hang on to your self: Of bodies, embodiment, and selves, Wolputte [/bib_ref]. However, [bib_ref] Led (astray) by genetic maps: the cartography of the human genome and..., Lippman [/bib_ref] "genetic imaginary" thesis warned how genetic explanations would become dominant to our future understandings of human health and behavior in society, noting ethical and eugenic concerns. The 1990s also saw the rise of surveillance medicine [bib_ref] The rise of surveillance medicine, Armstrong [/bib_ref] , the idea of genes as embodied risks [bib_ref] Doing the right thing: genetic risk and responsibility, Hallowell [/bib_ref] and new forms of somatic citizenship that were based on genetic understandings of the self [bib_ref] Genetic risk and the birth of the somatic individual, Novas [/bib_ref].
These ideas built oninsights that argued that new forms of social relations based on biology would form in terms of "biosociality" or biosocial identities of groups of patients. Near the end of the Human Genome project, there was a lot of interest in its' implications, for instance, in prospects of gene therapy for cures of rare diseases [bib_ref] Waiting for the cure: mapping the social relations of human gene therapy..., Stockdale [/bib_ref]. However, there were criticisms too as [bib_ref] Chronic illness as biographical disruption or biographical disruption as chronic illness? Reflections..., Williams [/bib_ref] explained how such genetic explanations of self and chronicity of the body were problematized and pathologised. We also noted a reaction in sociological literature against "speculative" scientific cures in the popular press, and warnings about "hype" and economy of raising false hopes [bib_ref] Hope against hype-accountability in biopasts, presents and futures, Brown [/bib_ref]. The sociology of disability also elucidated the difference between biological basis of impairment and experience of disablement by society, noting the social construction of "disability"or how biological forms of impairments could lead to social disablement. The best version of the self was biologically "healthy" andargued that an "imperative of health" was established -of not needing curative intervention -which became embedded in our social and cultural norms. described how this had progressed toward a "genetic imperative -the drive to find biological but above all genetic underpinnings for all things human, in sickness or in health". Within the literature and society in general, there was a continued engagement with geneticization in the 2000s which heralded more uncertainty linked to the increase of genetic explanations in medical, ethical and social discourses [bib_ref] Expansion and uncertainty: cystic fibrosis, classification and genetics, Hedgecoe [/bib_ref]. Lay perspectives became gendered as women had to become "moral pioneers", through engagement with increasingly invasive technologies of screening in antenatal care, genetic testing normalized and reproduction stratified in terms of inequalities. There were also calls for a sociology of health, illness and disease, to examine how biology and disease were intertwined [bib_ref] Body work in assisted conception: exploring public and private settings, Kerr [/bib_ref]. Despite warnings about the dominance of the genetic imaginary, [bib_ref] Have we seen the geneticisation of society? Expectations and evidence, Weiner [/bib_ref] argued that genetic explanations did not impact day to day life as much as would have been expected but also that those genetic imaginaries were now changing into molecular or pharmaceutical explanations.
Genetic understandings of embodiment also gained more complexity with epigenetic explanations, as there was not always a clearly defined path identified between disease and genetic expression of an illness [bib_ref] Sequencing and its consequences: path dependence and the relationships between genetics and..., Shostak [/bib_ref]. [bib_ref] Some reflections on the socio-cultural and bioscientific limits of bodily integrity, Shildrick [/bib_ref] argued that neither biomedical nor socio-cultural explanations of a bounded body seemed to hold sway. Yet, the clinical diagnosis of disease and biological understandings remained important to lay people, and patients demanded to talk about their experiences of diagnosis and we saw the beginnings of a sociology of diagnosis [bib_ref] I just want permission to be ill': towards a sociology of medically..., Nettleton [/bib_ref]. As genes became expressions of neoliberal risks and parts of the body commodified [bib_ref] The commodification of the body and its parts, Sharp [/bib_ref] , research began to examine ever smaller biological units of bodily consumption and literature on the bioeconomy developed [bib_ref] Theorizing the bioeconomy: biovalue, biocapital, bioeconomics or, Birch [/bib_ref]. Bioeconomies were associated with the rise of genomics and epigenetic advances which also entailed a paradigm shift in biomedicine away from genes to molecular biology and environmental determinants. According to Arribas-Ayllon (2016) the way in which to think of how this worked was not a constructionist approach but rather that of understanding "assemblages" of "genes, people and environments".
Ontologies and epistemologies have typically focused on the rise of biological realism and materialisms, in terms of genetic diagnosis and linked technologies such as, for example, the CRSPR-Cas9 gene editing technique. Thus, [bib_ref] Comprehending the body in the era of the epigenome, Lock [/bib_ref] warned against the collapse of the nature and culture divide in epigenetic explanations and rise of neo-reductionism or reduction to biology. However, it was argued that while health is becoming increasingly biomedicalized, rather than a vision of biological essentialism, research revealed shifting conceptions of identity but also wider biological understandings of the body as assemblages of (micro) chimeras, bacteria, viruses and even food [bib_ref] The chimera of liberal individualism: how cells became selves in human clinical..., Martin [/bib_ref] [bib_ref] Some reflections on the socio-cultural and bioscientific limits of bodily integrity, Shildrick [/bib_ref] [bib_ref] Food as exposure: Nutritional epigenetics and the new metabolism, Landecker [/bib_ref] [bib_ref] Blood functions: disability, biosociality and facts of the body, Fritsch [/bib_ref] [bib_ref] Marked 'h'for human: chimeric life and the politics of the human, Hinterberger [/bib_ref]. These unsettled the epistemological and ontological lines between human and non-human and how we understood who we were and in turn, what rights and responsibilities we accord ourselves and the environment. Moreover, it was becoming increasingly difficult to think humans without non-human symbionts (See, with arguments advocating understanding epigenetics, impact of intergenerational historical trauma on biology and future outcomes, in what Ingold and Palsson (2013) term "biosocial becomings" -to illustrate how the biological and the social, or nature and culture, are tied together. Yet, such arguments were often not reflected in the way in which personalized medicine or health surveillance was understood in practice.
Health surveillance became more encompassing, for instance, through (bio) digitalization and quantification of health in almost every aspect of daily life, as well as social understandings with respect to digital technologies [bib_ref] M-health and health promotion: the digital cyborg and surveillance society, Lupton [/bib_ref]. Classifications of illnesses, disorders, diseases and disabilities require differing forms of bio-technological diagnosis and clinical identification or biocertification [bib_ref] Blood functions: disability, biosociality and facts of the body, Fritsch [/bib_ref] before a "cure" could be found. This has implications for identity and social relations, in that diagnosis or biocertification seemingly "fix" a biomedical identity or label it. In such a way, the cause and also the cure must be found in biological explanations, as we see, for example, in connections between genetic explanations for some psychiatric conditions (Arribas-Ayllon et al., 2019). Personalized medicine has meant a rethinking of disease taxonomies for those that are genomics based, but which has been difficult to implement in clinical practice due to uncertainties of genomic variation [bib_ref] Plastic diagnostics: the remaking of disease and evidence in personalized medicine, Green [/bib_ref] [bib_ref] The rare and the common: scale and the genetic imaginary in Alzheimer's..., Milne [/bib_ref]. Despite this, new forms of biosociality are thus developing around more specific forms of biological data, toward biosocial potentiality, like the routine collection of biomarkers in social surveys as "precursors" of disability [bib_ref] Biomarkers as precursors of disability, Davillas [/bib_ref] or "exposomic" biological monitoring for environmental exposures [bib_ref] Biomonitoring in the era of the exposome, Dennis [/bib_ref].
Thus, [bib_ref] Normality and pathology in a biomedical age, Rose [/bib_ref] argues that we are witnessing a new form of active "global genomic citizenship" which accepts neoliberal forms of individual consumer responsibility for understanding risks of disease in our genomic data. Aligned is the idea that pathologies in genomic data is the new norm [bib_ref] Normality and pathology in a biomedical age, Rose [/bib_ref] and we are all genetic carriers of various risks. We now have to work at being and acting healthy which can involve: consuming certain foods; vitamins or superfood powder; exercise, medication; undertake various types of screening; engage in stress free social relations; as well as the routine collection of differing forms of genealogical and lifestyle data. An associated issue is this norm of hyper-engaged health, assumes that there is a hierarchy of health and able body, which we are and can be individually responsible for to prevent health risks. In such a way, the imperative of health, or norm of caring for ourselves by undertaking healthy "choices", is also increasingly becoming an imperative of curing oneself and a new form of individual and collective public health and "body work" [bib_ref] Body work in assisted conception: exploring public and private settings, Kerr [/bib_ref]. This kind of "body work" has an enduring correlation to health -to keep "fit" and "fitness" as industryas well as reproduction and fertility. For example, claims become correlated to perceived "fitness" to be able to mother and father a child [bib_ref] Fit to father? Online accounts of lifestyle changes and help-seeking on a..., Hanna [/bib_ref] , which is also increasingly about genetic future fitness and social potentialities (Arribas-Ayllon et al., 2019).
In the brief overview above, there is a tension between the indeterminacy and ever complexity of understandings of human selves and how curative potentials become implicated. This partly explains why we do not focus on "cure" as a general concept, as research on diagnosis, personalized medicine and what happens to identity or in the clinic while cure takes place are prioritized. However, this research occurred before the COVID-19 pandemic and in what follows, three examples are given to illustrate the logic of cure as it functions today and what that reveals. We look at care as cure, cure as care and cure as cost (see [fig_ref] FIGURE 1 |: The logic of cure. [/fig_ref].
## Case-study 1: disability and the pandemic: when care means cure
What does a neoliberal responsibility to cure oneself look like and is it always possible in practice? Discourses linked to fitness, cure and medicine, as well as contestations to genomic identity are not viewed neutrally by disabled people, those with chronic illnesses nor their organizations, who are often critical of technological advances to "cure" all [bib_ref] Disability activisms: social model stalwarts and biological citizens, Hughes [/bib_ref] but can also biosocially ascribe to medical identities or impairments due to the hopes of cure [bib_ref] Disability and chronic illness, Ottosdottir [/bib_ref] [bib_ref] Biosocial model of disability, Berghs [/bib_ref]. Within disability community, alongside such complexities, there have also been longstanding concerns about soft eugenics in screening technologies to prevent disability, the ideology of "ableism" or having to conform to ableist physical norms of health, and violence of medical cure and "curative imaginary" in terms of threat to the integrity of the body, "crip time" and disability identity.
Disabled people and disability activists often argue against (bio) medical or genetic understandings of disability, or biological essentialism and reductionism, instead pointing to social barriers to inclusion and equality, in keeping with social becomings or a social model of disability. The social model of disability makes a distinction between biology of impairment and disabling experience of social oppression through negative social attitudes and/or environmental barriers in society [bib_ref] The social model of disability: thirty years on, Oliver [/bib_ref]. It also draws attention to the inequalities across the lifecourse connected to both gaining impairment and experiencing disablement [bib_ref] The social model of disability: thirty years on, Oliver [/bib_ref]. An important aspect of such inequalities has always been a lack of access to health and social care. The social model of disability is also foundational to human rights approaches to disability which view access to health, and thus healthcare, as a human rights issue [bib_ref] Gene editing: heed disability views, Shakespeare [/bib_ref].
The COVID-19 pandemic illustrated many of the complexities of the relationship between disability and what a curative imperative now implies, in how healthcare access initially prioritized able-bodied health as norm in state policy [bib_ref] Disability, communication, and life itself in the COVID-19 pandemic, Goggin [/bib_ref]. As COVID-19 began affecting disabled people in care homes and locked institutions with discussions of medical rationing of vital equipment, such as oxygen and access to medicines in emergency situations, the focus turned to questioning medical ethics and (bio) ethical responses with utilitarianism seemingly trumping over not only the sanctity of life principles [bib_ref] People with disabilities in COVID-19: fixing our priorities, Sabatello [/bib_ref] , relational care but also rights based approaches to healthcare. The lack of access of basic health care was touted as disablist [bib_ref] Disability, disablism, and COVID-19 pandemic triage, Scully [/bib_ref] and the triaging or giving of priority to able-bodied and healthy people to receive therapies that could cure, ableism. Yet, at the heart of such policies was an understanding of "recovery" or who would be curable which is different from ableism, in that it is not just able body that is prioritized but temporal ability to stay well. Initially, policy responses benefited the able-bodied majority and while purporting to protect the "vulnerable", policies caused disabled people to begin to experience risks to COVID-19 [bib_ref] An observational cohort study of numbers and causes of preventable general hospital..., Glover [/bib_ref]. Policy decisions also lacked inclusion of disabled people, their organizations or basic accessibility, such as government announcements that were inaccessible, for example, by holding televised government broadcasts with no sign-language interpreters in the UK.
Due to inadequate Personal and Protective Equipment (PPE) in hospitals and residential facilities, lack of accessible information and the discharging of people from hospitals into care homes and residential facilities, particular disabled groups were more at risk for COVID-19 and dying early on in the pandemic. These included older people, those with co-morbidities and people with learning difficulties and autism. Those were disabled groups that were particularly discriminated against within health and social care systems and had been neglected pre-pandemic (Inclusion Europe, 2020). Intersectionality also began to affect who was most at risk from COVID-19 with inequalities revealed in gender, ethnicity, impairments, age and socio-economic status in who was impacted directly and indirectly. Despite COVID-19 being linked to inequalities and deprivation, (bio) medical model discourses around "underlying conditions", "genetic dispositions" or biological "racial" risks began to take precedence over explanations of structural disadvantage or syndemics [bib_ref] Race disparities in the COVID-19 pandemic-SOLUTIONS lie in policy, not biology, Boulware [/bib_ref] [bib_ref] Systemic racism, chronic health inequities, and COVID-19: a syndemic in the making?, Gravlee [/bib_ref].
Vulnerability to impairments and deaths that were created in the COVID-19 pandemic illustrated dangers of narrow understandings of cure and (bio)ethical responses that only focused on care or rights. Arguments about biological essentialism and dangers of bodily reductionism, despite epigenetic complexity of how we understand our "selves" sociologically and anthropologically, proved to be persuasive in practice. Likewise, the reductive genetic imaginary of people's biological potentialities or dispositions for cure instead of ethical inclusion and rights of equity of treatment [bib_ref] The COVID-19 response must be disability inclusive, Armitage [/bib_ref] , were foundational not only in popular discourses but in preventing access to medical treatment. Noteworthy, is the rise of an encompassing biosocial model of disability [bib_ref] Biosocial model of disability, Berghs [/bib_ref] that now affects everyone because it is predicated on epigenetic risk and biological impairment. In the next example, we look at the curative hopes of the general population and how inequalities become stratified in vaccine nationalism and priorities of vaccination during the COVID-19 pandemic.
## Case-study 2: learning to live with indeterminacies of cure: covid-19 vaccines and cure as care
Patients increasingly advocate for diagnosis, better care and also rights to cures [bib_ref] The political economy of hope: patients' organizations, science and biovalue, Novas [/bib_ref] and care, in what Jae (2018) terms "anticipatory politics". Curing is bound to the political economy of hope [bib_ref] The political economy of hope: patients' organizations, science and biovalue, Novas [/bib_ref] and hopelessness [bib_ref] Imagined futures in living with multiple conditions: positivity, relationality and hopelessness, Coyle [/bib_ref] , as well as transnational promises of the genomic technologies of cure and evidence-based health activism and experimentation [bib_ref] Evidence-based activism: patients' , users' and activists' groups in knowledge society, Rabeharisoa [/bib_ref]. Curing now also has an imperative of preservation or repairing of health as site of norm of life which is why "getting back to normal" featured in popular discourses around the race for vaccines for the SARS-CoV-2 virus that causes COVID-19. Despite sociological warnings of "hype" or anticipation of new treatments and unfulfilled promises in expectations of normalcy [bib_ref] Hope against hype-accountability in biopasts, presents and futures, Brown [/bib_ref] , hopes of new forms of treatment that will cure or prevent illness will dominate patient, public and technological discourses of innovation. Actions, emotions as well as technological developments for cures are temporally directed toward "promissory futures" [bib_ref] Hope against hype-accountability in biopasts, presents and futures, Brown [/bib_ref] , "curative imaginaries"or "imagined" futures free of viruses (Van Loon, 2013), which a global community is politically, social, culturally and financially invested in. However, such speculative futures become problematic if the general publics' conceptions of cure do not tie in with indeterminacies on offer.
Mass vaccination programmes are a given in many countries in the Global North with Pfizer/BioNtech, AstraZeneca/Oxford, Janssen/Ad26.COV 2.S, Moderna mRNA and Sinovac-Corona Vac among the vaccines being given to populations. However, while most of these vaccines could not offer absolute cures, they seemed to be presented in terms of reduction of total risks and representing individual and national forms of viral sovereignty linked to state political and economic interests [bib_ref] Virotech patents, viropiracy, and viral sovereignty, Yu [/bib_ref]. Within many of the COVID-19 pandemic vaccination campaigns, it is clear that cure is still understood as "absolute" for the general population. This meant offers of multiple vaccine doses and promises of getting back to a normal that did not initially materialize, were treated with suspicion, hesitancy and disbelief. Uptake of vaccines were also dependent on how cure was linked to experiential understandings of other illnesses and viruses and their risks along an affective and temporal pandemic continuum that ebbed and flowed [bib_ref] Associations of COVID-19 risk perception with vaccine hesitancy over time for Italian..., Caserotti [/bib_ref]. The risks and fears of safety of vaccines expressed by populations also often implicated many of the changing conceptions of the self and embodiment found in sociological and anthropological research, but this was generally ignored or just seen as misinformation about vaccines [bib_ref] Understanding COVID-19 misinformation and vaccine hesitancy in context: findings from a qualitative..., Lockyer [/bib_ref]. For instance, some refusals of vaccines encompassed ideas implicit in healthism, the body as being fit, having immunity and being able to "cure" itself which have been very popular in the health, wellness and fitness industries used to sell products to consumers.
While citizenship and risks were connected to neoliberal individual responsibilities for consumer choice, there were no real informed choices over vaccine uptake, as people could not weigh individual experiential risks, for instance of sideeffects, impact on co-morbidities or other conditions, nor what level of immunity a vaccine would give, by choosing which vaccine to have. Instead appeals were made to engagement of cure as care and ethical responsibilities to protect fellow citizens. Those kinds of sentiments while important, might be at odds with neoliberalism (or even libertarianism) focused on individual and consumer choice. Similarly, such appeals may not work for populations who have experienced historic and present inequalities of care, lack trust in their governments and/or suffered abuses from the medical establishment. While wanting to care for the self and others, some communities may also feel as if they are taking part in an experiment, feel hesitant about lack of informed choice or want more evidence of viral effects before taking up the vaccine offer [bib_ref] Hopes, hesitancy and the risky business of vaccine development, Calnan [/bib_ref] [bib_ref] Understanding COVID-19 misinformation and vaccine hesitancy in context: findings from a qualitative..., Lockyer [/bib_ref]. This is normal when none of the differing vaccines could promise a hundred percent immunity to the SARS-CoV-2, with reports in the press and on social media often weighing vaccines up against each other, leading to popular reports of vaccine envy, dissatisfaction and regret over lack of choices. It is crucial to note that this is in the context of vaccines presented as "cures" with a history of economic and social sacrifices as well as deaths from COVID-19. There was also discussion of long COVID in the popular press, as well as possible future of COVID-19 variants. In such a context, it seems entirely reasonable that people would want and need to make an informed choice about choice of vaccines and this could be emotional. This also changes the way in which public health measures are viewed because if we have vaccines presented as "cures" then that must mean that the collective body and state are "back to normal" and health? Even if vaccines are understood, the idea of "herd immunity" might also mean that certain public health measures are no longer acceptable, regardless of what the evidence or politicians might say.
Undoubtedly, while many people were also grateful and happy to take up vaccine offers as responsible citizens, there was ambivalence toward not only indeterminancies of cure presented but also with regards to local, national and international inequalities and curative ethics [bib_ref] Understanding COVID-19 misinformation and vaccine hesitancy in context: findings from a qualitative..., Lockyer [/bib_ref]. All over the world people had died or had gained impairments, and in the Global North, consistent with structural disadvantage and inequalities, it was particularly ethnic minority communities that had been badly affected by COVID-19 [bib_ref] Understanding COVID-19 misinformation and vaccine hesitancy in context: findings from a qualitative..., Lockyer [/bib_ref] , which led to the pandemic being understood as racialised. It was also mainly richer countries in the Global North, like the UK and United States (US), that engaged in vaccine nationalism rather than vaccine equity [bib_ref] From vaccine nationalism to vaccine equity-finding a path forward, Katz [/bib_ref] , often prioritizing national interests over global wellbeing. In countries like Brazil and India, there have also been accusations of curative fascism in lack of national planning to ensure that the population could get access to vaccines and have rights to cure. Within a global curative imaginary, access to vaccines represented survival not the "getting back to normal" of the Global North. While there are programmes such as COVID-19 Vaccines Global Access (COVAX) to ensure equitable global vaccine distribution, directed by GAVI, the Vaccine Alliance, the Coalition for Epidemic Preparedness Innovations, and the World Health Organisation (WHO), such international partnerships seemed to have to sucede urgency to national priorities which also protected pharmaceutical investments and distributions in the Global North [bib_ref] The role of good governance in the race for global vaccination during..., Tatar [/bib_ref].
This case-study reveals how national interests and short term visions of quick fixes and partial curative hypes have a stronger hold than human rights to health, global ethics of care and international treaties or agreements. Such short-term visions when it comes to cure and care are also dangerous, not only in the affective experience of ambivalence over having vaccines, public health measures not seen as needed, as well as leading to development of new variants of the COVID-19 virus that might be more virulent and possibly evade the vaccines that have been developed. To date there have been several variants but none that have evaded vaccine development. However, for instance, both Delta and Omicron have been very transmissible and lead to hospitalizations of unvaccinated people and some vulnerable members of the population. This has brought ethical questions to the fore in how far people's freedoms not to choose to be vaccinated or cured can be understood, especially if it could lead to their hospitalization, them gaining impairments or even causing their deaths or those of more vulnerable members of a community [bib_ref] Winter of omicron-the evolving COVID-19 pandemic, Del Rio [/bib_ref]. As such, some workplaces and countries have insisted on vaccines as mandatory, especially as certain public health measures become unacceptable and risks of transmission can no longer be contained but also may not need to be feared with vaccination.
As such, the vaccine response, so far, illustrates the limitations of curative hopes and how important it is to be transparent and explain what kind of cure is on offer, for example, that it will be partial, need top-ups and is not absolute. It also takes seriously that populations may have sophisticated ideas about how the biological and human environment become implicated in cures and understandings of viruses. This entails that discourses around COVID-19 should now include a new normal where learning to live with increased risks and having to top up vaccines each year is part of curative trajectories. This would give strength to understanding why cure as responsible act of citizenship and care for each other then becomes important. However, what is missing is an understanding of how unequal participation in cures ties into ethical ambivalences around racialised inequalities.
## Case-study 3: rare diseases and weighing the costs of therapies or cure: choosing cures -business as normal?
It seems as if cures initially mainly benefit certain populations in the Global North but is this always the case? The transnational nature of seeking cures in both proven and unproven treatments as well-curative experiments is only expected to grow with the "genetic imperative" [bib_ref] Genetics, biosocial groups & the future of identity, Hacking [/bib_ref] and will lead to difficult decisions for disease and patient organizations and community groups, in terms of understanding and delineating between evidence of treatments or trials that are legitimate and regulated; and those that do not work or are speculative [bib_ref] Between hope and evidence: how community advisors demarcate the boundary between legitimate..., Petersen [/bib_ref]. [bib_ref] The anticipatory politics of improving childhood survival for sickle cell disease, Jae [/bib_ref] argues that "structural conditions also stratify expectations for the future, including the affective appeal of medical innovations" concerning what treatments or trials for cure are available and what rate of success they have. [bib_ref] Genomics and cure: understanding narratives of patients with Duchenne muscular dystrophy in..., Kato [/bib_ref] too explains that while genomics becomes tied to cure, nuances have to be made in how patients understand and want cures, with regards to present histories of discrimination, impact of racism, lack of care and scientific understanding about genomic medicine. In many ways, a new language of choices of cures and therapies that takes into account such concerns is needed and developing.
If we take "rare" diseases, such as, for example the inherited genetic blood disorders sickle cell and thalassaemia that affect millions of people worldwide, recent innovations in terms of therapies and curative genomic realities in gene editing [bib_ref] CRISPR-Cas9 gene editing for sickle cell disease and β-thalassemia, Frangoul [/bib_ref] with "functional cures", convey numerous difficulties in a global transnational context when we think of hopes of cure. While these conditions affect millions worldwide, in the UK they mainly affect ethnic minority populations, thus national classifications of how "rare" the respective conditions are and their applications, or scalability of those innovative cures for other common conditions, are key [bib_ref] The rare and the common: scale and the genetic imaginary in Alzheimer's..., Milne [/bib_ref]. Those considerations around therapies and curative realities also occur against the background of history of curative neglect and racialised pandemic where people with such rare diseases are identified as "clinically extremely vulnerable" (CEV) and have been encouraged to get vaccinated because they are at higher risks of morbidity and mortality due to their conditions and structural inequalities affecting them as minorities. What does cure mean against such a background and how are the costs and risks of cure weighed up and by whom? What does that mean for a patient's individual choice of cures and therapies?
In the UK, screening services in antenatal care, genetic testing, treatments such as hydroxyurea for sickle cell, blood transfusions for both sickle cell and thalassaemia, specialized clinical care and public advocacy have entailed that the conditions can now be viewed as chronic and long-term, whilst in many other places in the world, without access to adequate healthcare, they can be acute and disabling. In the Global North, innovative genomic developments have entailed a shift of research attention so it is focused on improvements in access to therapies, as well as curative possibilities for those with differing expressions of these genetic disorders [bib_ref] The anticipatory politics of improving childhood survival for sickle cell disease, Jae [/bib_ref]. Stem cell transplants for children with the most serious clinical manifestations of the conditions, often from a genetically matched relation such as donor sibling, is an option, in terms of cure, but was not without serious risks, complications and mistrust. This has often raised difficult ethical, social and emotional issues about how to make and share these decisions within families [bib_ref] Shared decision making in hematopoietic stem cell transplantation for sickle cell disease, Sullivan [/bib_ref] and what the rights are of older patients facing more serious manifestations of the conditions to such risky or newer experimental cures. A politics of social justice, "race" and health equity in terms of access to cures in resource poor settings in the UK, US and countries in the Global South where most people are affected; like in Nigeria, India and Brazil, that do not have access to the basics of good care, has also emerged [bib_ref] Time to apply a social determinants of health lens to addressing sickle..., Berghs [/bib_ref].
The experimental possibilities and understandings of cure have changed with the advent of genomics and personalized medicine, making real immunotherapeutic possibilities such as gene editing out the mutation, stem cell transplantations in adults without need for chemotherapy or related donors, and gene therapy to change the affected genes [bib_ref] Risk and Technological Culture: Towards a Sociology of Virulence, Urnov [/bib_ref]. Many of these experimental possibilities of cure are being tested in countries in the Global North in randomized controlled trials (RCTs) and have reversed the trajectory of ethical, relational and social dilemmas, for example, with respect to questions about if children should also have access to these experimental treatments and how decisionmaking should then occur [bib_ref] Are the risks of treatment to cure a child with severe sickle..., De Montalembert [/bib_ref] and finances be found [bib_ref] Risk and Technological Culture: Towards a Sociology of Virulence, Urnov [/bib_ref]. Alongside such experimental treatments for long neglected conditions, are hopes invested in new forms of therapies and medicines that have been developing that might mean less curative risks and increasingly manageable long-term conditions. For instance, Crizanlizumab has recently been approved as a new therapeutic treatment in the UK for patients with sickle cell. This is the first one in over 20 years and further therapies are in development and undergoing trials. This raises the critical question of who "curates" or who gives access to cures and why? And can patients have choices between cures and therapies? (see [fig_ref] FIGURE 2 |: Simplified version of therapies and curative options for sickle cell condition [/fig_ref].
For instance, in the UK recently, the National Institute for Health and Care Excellence (NICE) has not recommended the use of Bluebird Bios' Zynteglo (betibeglogene autotemcel) gene therapy for the treatment of patients with transfusion dependent beta-thalassaemia and is undergoing its appraisal. The national, and in particular the international patient support group, Thalassaemia International Federation (TIF, 2021), quickly reacted against the decision, noting how this was based on an understanding of thalassaemia as a manageable chronic condition with a good quality of life and life expectancy, which ignored the day to day difficulties and accumulative impact of the condition on impairment. Additionally, they argued that NICE weighed the short and especially long-term risks to the relatively small cohort of patients that would benefit, as well as cost-effectiveness of such a treatment to the NHS but TIF (2021) noted that for some patients those long-term risks of an uncertain therapy would be justified and that NHS estimations of cost-effectiveness did not take into account that this condition affected a minority ethnic population group with greater need for equity. Among the short term risks that NICE had to consider were that the same gene therapy, formerly called Lentiglobin, was also being tested in patients with sickle cell but due to two serious adverse reactions in patients, they had to pause their clinical trials [bib_ref] Bluebird bio halts sickle cell trials after leukemia diagnosis, Terry [/bib_ref].
However, the US Food and Drug Administration (FDA) deemed those adverse serious reactions were unrelated to the gene therapies and clinical trials with sickle cell patients resumed [bib_ref] After analysis, bluebird bio says vector "very unlikely" cause of acute myeloid..., Philippidis [/bib_ref] and the gene therapy was put on the UK's Medicines and Healthcare Products Regulatory Agency's new 2021 Innovative Licensing and Access Pathway to ensure quicker access for patients to medicines but seemingly only impacting sickle cell patients (a relatively larger ethnic minority group) and not those with beta-thalassaemia where NICE is still assessing the treatment. By contrast, the European Union's Medicines Agency's Pharmacovigilance Risk Assessment Committee (PRAC) approved Zynteglo, deciding that there was a favorable benefits and risks profile and discounting some of the earlier risks presented in early trial stages. This is still not without risks for certain patients (if they meet the conditions set out for use), as well as if conditionalities of insurance companies financing are met in the successful eradication of main symptoms of being transfusion dependent in those betathalassaemia patients [bib_ref] Risk and Technological Culture: Towards a Sociology of Virulence, Urnov [/bib_ref]. However, there is also a wider perspective in understanding the interest in regulatory approvals, as such "rare" conditions represent the experimental innovative firsts in curative potentialities to be scaled up to other conditions. That is why despite a pandemic, there are political, social and financial interests in such innovation, how patients will access them and what implications they will have.
The above illustrates how local and global curative hopes are tied together but access to cures are curated by national and international bodies, patient advocacy groups as well as private and public partnerships, weighing up risks, ethics and understandings of quality of life, as well as future financial rewards to health systems by genomic innovations. However, cures are "functional" meaning that symptoms are controlled or eradicated but people still carry the gene which, for instance, has reproductive, social, cultural and psychological implications. Likewise, while treatments like gene editing represent a curative revolution, we do not know much how patients involved in those trials and treatments experience cure and describe it nor when such innovations will benefit those in the Global South. In that sense, we are also at the start of a revolution of genomic cures and their unequal impacts on differing groups of patients and communities.
# Discussion
We do not directly investigate cure itself nor describe it in general terms in sociology and anthropology, as research on diagnosis, personalized medicine and what happens to identity, through patient advocacy, in a lab or in the clinic while cure takes place or in prognosis, are prioritized. I have taken a step back from the precision evident in the vast literature by noting how an omission develops around "cure" and what it means empirically to undergo differing forms of cures. In the above, I have taken cure seriously as a concept to be investigated and have outlined how a logic of cure operates through three short case studies. It was noted how the act of curation or sorting through who or what will be cured or how certain diseases, disabilities or illnesses become curable or incurable biomedically, for national state interests or due to regulatory bodies responsible for assessing risks and financial rewards of novel cures, ignores many of developments found in the sociological and anthropological literature. Instead, the logic of cure is an imperative and commodified according to a biopolitics of differing forms of cure; from the seemingly incurable, to partial cures and the functionally curable. A new language around cures begins to develop and reveals what cures mean and could represent socially, affectively and temporally to patients and broader society. However, it is clear too that the language and theories of ethics and rights are critically missing in how we describe cures. We noted this was being contested.
Within disability studies and disability community responses to impact of the COVID-19 pandemic, there was the development of a "bioethics of cure" [bib_ref] Who gets cured? COVID-19 and developing a critical medical sociology and anthropology..., Berghs [/bib_ref] in criticism of: (1) lack of care as cure; (2) the ableism evident in who got access to care and why; and (3) policy creation of "vulnerability" to impairment and disablement of COVID-19. Within vaccine hesitancy we also noted ambivalence between appeals to neoliberal citizenship and responsibilities of care within a racialised pandemic that was correlated to inequalities of society. Thus, a bioethics of cure is also informed experientially by a lack of local and global ethics of care and indeterminacies of cure. [bib_ref] Embodied risk: my body, myself?, Kavanagh [/bib_ref] emphasized that if you wanted to understand intersection between environmental and embodied risks, it was important to work together with people at "risk" to formulate new languages as well as approaches to those environmental and socially embodied understandings. Yet, a bioethics of cure is an empirical-ethical theory that is developing from the experiential knowledge of people who do not have access to differing forms of cures [bib_ref] The experiential knowledge of patients: a new resource for biomedical research?, Caron-Flinterman [/bib_ref] , and are undertaking differing forms of "curative labour"or curative risk work to stay well and protect themselves. It is also still developing socially and will have to address the rights to impairment, not to know biosocial potentiality of cure, as well as demonstrate more ethical-rights based responses to, for instance, some people's (initial) vaccine hesitancy as choice or even total curative refusals.
Epistemological and ontological understandings of embodiment during COVID-19 were also being influenced by a biological neo-reductionism. That neo-reductionism of people to their biological immune responses, influenced medical and policy understandings of who has potentiality for COVID-19, as well as could be cured, showed an urgent need for sociological and anthropological engagement with cure. The latest developments in biosociality note that we will all have several illnesses, potentialities for disabilities and risks for conditions across the life course and these are also affected by epigenetics, inequalities and intersectional identities. However, instead of such nuanced understandings of embodiment affecting the way in which vaccines were presented, they were understood as absolute cures rather than partial with need for top-ups to ensure robust immune responses. The public had sophisticated understandings of vaccines and viruses that were not being matched by explanations of cures. Likewise, there was an ambivalence around engaging in cures that were unequal and not being given to everyone, with dangerous possible consequences.
Nowhere was this ethical ambivalence illustrated more than in the latest genomic advances, like gene therapy, that are now offering real possible cures for rare genetic conditions but trialed on populations in the Global North with strong curative advocacy groups [bib_ref] Going for the cure: patient interest groups and health advocacy in the..., Keller [/bib_ref]. Those cures and access to them, are carefully nationally curated and regulated, and will not immediately benefit the Global South where most people with the conditions of sickle cell and thalassaemia are located. The emphasis there has also been on private and public partnerships with financial rewards of curing a rare disease rather than on what that entails socially, culturally, ethically, financially and biologically for patients and their families. While cure is now being described and viewed as "functional", it is still seen in isolation from a life-course perspective where there will be numerous illnesses, diseases and conditions needing care, and where you live in the world will affect how you understand the impact of what a cure or even therapy could have. [bib_ref] Genetics, biosocial groups & the future of identity, Hacking [/bib_ref] stated that we would have people coming together in terms of new types of genetic risks, and I argue that we will have new forms of identity emerging in terms of curative potentials and how those unsettle epistemologies and ontologies of the body, identity, embodiment and environment.
# Conclusion
In this article, I have set out the beginnings of a sociological and anthropological exploration of curative imperative as new normal, as well as bioethics of cure which has not been adequately examined in the literature despite understandings of how cure is changing. I have elucidated some key conceptual ideas, for example, that there is a logic of cure, connected to why epistemological and ontological positions are now at stake which debunk ideas about "biological determinism" and introduce complexities into former absolute ideas of cure toward understandings of curing. I argued that we have to examine what undergoing cure means and how hopes are invested in new technologies of cure but also understand concerns of patients or entire populations before, during and after they undergo different types of "cure" or are being "curated" as lacking certain curative potentialities. Throughout the paper, I have pointed to how cure links into categories of intersectionality that are not fixed biologically, and why they are stake in terms of how care and cure are connected to each other in the Global North and South.
# Author contributions
The author confirms being the sole contributor of this work and has approved it for publication.
[fig] FIGURE 1 |: The logic of cure. [/fig]
[fig] FIGURE 2 |: Simplified version of therapies and curative options for sickle cell condition. [/fig]
|
EGFR Mutant Structural Database: computationally predicted 3D structures and the corresponding binding free energies with gefitinib and erlotinib
Background: Epidermal growth factor receptor (EGFR) mutation-induced drug resistance has caused great difficulties in the treatment of non-small-cell lung cancer (NSCLC). However, structural information is available for just a few EGFR mutants. In this study, we created an EGFR Mutant Structural Database (freely available at http://bcc.ee.cityu.edu.hk/ data/EGFR.html), including the 3D EGFR mutant structures and their corresponding binding free energies with two commonly used inhibitors (gefitinib and erlotinib). Results: We collected the information of 942 NSCLC patients belonging to 112 mutation types. These mutation types are divided into five groups (insertion, deletion, duplication, modification and substitution), and substitution accounts for 61.61% of the mutation types and 54.14% of all the patients. Among all the 942 patients, 388 cases experienced a mutation at residue site 858 with leucine replaced by arginine (L858R), making it the most common mutation type. Moreover, 36 (32.Figure 6 Distribution of binding free energies of all the EGFR mutants with (A) gefitinib and (B) erlotinib. The energy components and the total energies are shown in the diagrams.
# Background
As the primary type of lung cancer, non-small-cell lung cancer (NSCLC) has received growing attention from the researchers. It is reported that about 85% of all the lung cancer patients are diagnosed as NSCLC. One strategy commonly used in the treatment is to target the tyrosine kinase (TK) domain of epidermal growth factor receptor (EGFR) to interrupt the downstream signaling. Reversible tyrosine kinase inhibitors (TKIs), such as gefitinib and erlotinib, are generally applied in this procedure. They are proven to be efficient for patients over a period of time, but a limited treatment outcome usually occurs because of mutation at EGFR TK domain. According to statistics, about 10% to 15% of white patients and 30% East Asian patients experience a mutation of EGFR TK domain, and over one hundred mutation types have been found so far.
Structural information is available for just a few EGFR mutants from the Protein Data Bank (PDB). They are obtained with experimental methods, such as X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy. These methods can produce highresolution protein crystal structures, but they are usually very complex, costly and time consuming. Bioinformatics based methods have become very popular and successful in predicting protein structures. Wang et al.predicted EGFR mutant structures using the tools scap and loopy. Yarov-Yarovoy et al.employed Rosettato predict helical transmembrane protein structures. The binding free energy acts as a useful index to evaluate the binding affinity between mutants and drugs, and can be used as an important indicator of drug resistance. Zhou et al.predicted EGFR mutation induced drug resistance based on the binding free energy, which was calculated with Amber. As different mutations affect the EGFR structure and drug resistance level differently, a database of the EGFR mutant structures and the corresponding binding free energies with TKIs can provide a useful resource for further research and clinical guidance.
In this study, we created an EGFR Mutant Structural Database, containing over one hundred EGFR mutants and their binding free energies with reversible TKIs gefitinib and erlotinib. We employed Rosettato generate the 3D structures of the EGFR mutants with the wild-type (WT) EGFR. Then Amberwas used to optimize the structures and compute the binding free energies with gefitinib and erlotinib. The procedure we have used to build the database is shown in .
# Methods
## Data collection
The EGFR mutation types were obtained from the EGFR Mutation Database (http://www.cityofhope.org/egfr-mutation-database)and the Queen Mary Hospital in Hong Kong. The EGFR Mutation Database is a public database, while the data from Queen Mary Hospital in Hong Kong were obtained through several clinical projects and all of these projects had ethics approvals. None of any data entries contains patient identity. Before the commencement of this study, we obtained approval and permission from Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster to use the data from Queen Mary Hospital. The mutations locate at exons 18 to 21 of the EGFR TK Procedure used to build the EGFR Mutant Structural Database. First, we applied Rosetta ddg_monomer protocol and comparative modeling (CM) protocol to predict EGFR mutant structures. Secondly, the predicted structures were refined with a minimization step using sander in Amber. Then a drug (gefitinib or erlotinib) was added to the mutant structures followed by MD simulation. Subsequently, we employed MM-GBSA in Amber to calculate the binding free energies of the EGFR mutants and the inhibitor. Finally, the refined mutant structures and their corresponding binding free energies with gefitinib and erlotinib were collected to establish the database.
domain. Specifically, there are 112 mutation types, including 95 from 774 NSCLC patients of the Mutation Database and 17 from 168 patients of the Queen Mary Hospital in Hong Kong. These mutation types are named according to their corresponding changes of the amino acid sequences relative to the WT sequence. In the mutation representation, A and B, locating at positions p and q respectively, represent two residues in the protein sequence. I is a single residue or a residue list, and C and D are two other residues.
The crystal structures of EGFR mutants with L858R and G719S are available from PDB. Other EGFR mutants, used for calculating binding free energies with gefitinib and erlotinib, were generated based on the template structures "2ITY" and "1 M17" respectively downloaded from PDB.
## Point mutation modeling
Residue substitution accounts for more than 60%of all the mutation types in our database. Single-point mutation is a replacement of an amino acid in the protein sequence with another, while double-point mutation occurs when the amino acids at two positions are replaced. In this paper, ddg_monomer in Rosetta was employed to generate the residue substitution mutants. This application takes the WT EGFR crystal structure and the mutant sequence as inputs, and the output is the structural model of the mutation with the side-chain replaced. Two main protocols are available in this procedure, the high-resolution protocoland the low-resolution protocol. The high-resolution protocol allows the backbone with a limited relaxation while the low-resolution protocol makes the backbone fixed. We adopted the high-resolution protocol to generate the EGFR single-point and doublepoint mutations. First, the side-chain at the mutant position is replaced and the Rosetta's standard side-chain optimization module is applied to optimize the rotamers at all residues of the protein. Subsequently, gradient-based minimization is adopted to produce the minimized structures. As the high-resolution protocol allows the backbone a limited freedom, C α -C α distance restraints are used in the optimization process in order to prevent the backbone moving too much from the start conformation.
## Homology modeling
We employed the homology modeling (also known as comparative modeling (CM))protocol in Rosetta to generate the mutations of amino acids insertion, deletion, duplication and modification relative to WT EGFR. Homology modeling is widely used in predicting protein structures as it can often provide reliable and accurate structural models. It provides a way to fill the large gap between the increasing number of available protein sequences and the protein crystal structures obtained from experimental methods.
Before model construction, several files (target and template protein sequences, template PDB file, mutantaligned sequences, fragment library and secondary structure file of the target) should be prepared first. Selection of a template is very important because it can affect the accuracy of the predicted structure. In our studies, the crystal structures of EGFR TK domain "2ITY" and "1 M17" are selected as templates to generate the mutants. After the template is determined, mutant sequences are aligned to the template sequence with multiple-sequence alignment program ClustalW. The fragment library includes short peptide backbone fragments, which can play an important role in the construction of variable regions. We employed the fragment picker protocol in Rosetta to pick fragments, which can help to establish models more efficiently and accurately by enabling rapid search of the conformational space. Moreover, PSIPREDare used to obtain the target's secondary structure file. After all these files prepared, the CM protocolin Rosetta are applied to build the well-aligned regions and the missing parts are rebuilt using loop modeling with the fragment library. Finally, a fullatom refinement step is performed to the models and clustering method is used to select models.
## Model assessment
The models predicted with software simulation may not be accurate, thus the verification and assessment of the predicted models become very important. Two methods are often adopted to assess the predicted models with software simulation, computing the energy of the model and evaluating the similarity with a given characteristic between the predicted model and the real structure. In this paper, we used physics-based energies of the predicted EGFR mutants to assess the accuracy of the 3D structures. The full atom energy scoring function was employed to calculate the energies of all the structures and the one with the minimum energy was identified as the finally predicted structure. Using the function, each predicted structure is scored with a series of parameters (Lennard-Jones interactions, solvation, residue pair interactions, van der Waals, hydrogen bonding, Ramachandran torsion preferences, rotamer self-energy and unfolded state reference energy) and their corresponding weights. The total score of a predicted model is defined as the weighted sum of all the scoring parameters.
## Molecular dynamics (md) simulation
After the predicted EGFR mutant structures were obtained, we optimized these structures using MD simulation in Amber. The simulation is conducted in a solvent environment, thus, an octahedron water box (TIP3P model, 10.0-angstrom (Å)) is added to the structure with tleap in Amber. In order to describe the molecule interactions, the following molecular force field is adopted in Amber.
[formula] V r ð Þ ¼ E bonded þ E nonbonded ¼ X bonds K b b−b 0 ð Þ 2 þ X angles K θ θ−θ 0 ð Þ 2 þ X dihedrals V n =2 ð Þ 1 þ cos nϕ−δ ½ ð Þ þ X nonbij A ij =r 12 ij − B ij =r 6 ij þ q i q j =r ijð1Þ [/formula]
The total energy is composed of bonded term E bonded and non-bonded term E nonbonded . In Equation (1), the bonded energy which is related to the covalent bonds consists of bond stretching (where K b is an empirical stretching force constant, b and b 0 are the actual and empirical bond lengths respectively), angle bending (where K θ is a constant, θ and θ 0 are the actual and empirical bond angles respectively), and torsion terms (where V n is the barrier to free rotation for the empirical bond, n is rotation periodicity, ϕ stands for torsion angle, and δ represents the angle when the potential reaches its minimum value). The nonbonded energy includes van der Waals (where A ij and B ij describe the depth and position for a pair of non-bonded interacting atoms respectively, and r ij is the interatomic distance) and the long-range electrostatic terms (where q i and q j are point charges, and r ij is the interatomic distance). In our simulation, we employed the ff99SB force field, which is a broad application of the basic force field. After solvating the complex and adding force filed, we conducted a minimization step to the entire system with sander in Amber. The result from the optimization process is our refined mutant structure.
With MatchMaker in UCSF Chimera, we aligned the optimized structure to the template complex "2ITY" (EGFR-gefitinib complex) or "1 M17" (EGFR-erlotinib complex) to obtain the mutant-drug complex. Then Amber was used to optimize these complexes. Similarly, the complex was solvated in a TIP3P water box (10.0 Å) and the ff99SB force filed was adopted. In order to conduct the production MD, we need to equilibrate the solvated complex using sander in Amber. First, 1000 circles of minimization were adopted to remove any bad contacts and make the structure relaxed. In this procedure, steepest descent algorithm was used for the first 500 steps and conjugate gradient algorithm was applied for the second 500 steps. Then 50 picosecond (ps) of heating and 50 ps of density equilibration were conducted to reach the temperature about 300 K and the density around 1 grams/ml. Subsequently, equilibration of constant pressure at 500 ps was carried out at the temperature of 300 K. All these simulations were conducted with shake on hydrogen atoms, and Langevin dynamics was used to control the temperature. Several parameters, such as temperature, density, total energy and root-mean-square deviation (RMSD) were finally used to verify that the equilibration of the system. When the system is equilibrated, we proceeded to run the production MD for a total of 2 ns and recorded the coordinates every 10 ps.
## Binding free energy calculation
The binding free energy of each mutant-drug complex is calculated based on the motion trajectories, which are generated during the production MD simulation. The MM-GBSA method in Amber tools was applied to calculate the binding free energies of EGFR mutants and reversible TKIs (gefitinib and erlotinib). The aim of this procedure is to obtain the free energy difference between the bound and unbound state of two solvated molecules. However, in a solvent environment, the solvent-solvent interactions account for most energy contributions and the fluctuations of the total energy would be an order of magnitude greater than the binding energy. Therefore, the binding free energy is calculated as follows by means of thermodynamic cycle in solvent and vacuum environment.
[formula] ΔG bind;solv ¼ ΔG bind;vacuum þ ΔG solv;complex − ΔG solv;ligand þ ΔG solv;receptor À Áð2Þ [/formula]
where ΔG bind,solv and ΔG bind,vacuum represent the free energy difference of bound and unbound state of a complex in solvent and vacuum environment respectively, and ΔG solv,receptor , ΔG solv,ligand and ΔG solv,complex stand for the changes of free energies of the receptor, ligand and complex between solvent and vacuum environment, respectively.
We calculated the binding free energies of EGFR mutants with gefitinib and erlotinib. MM-GBSA in Amber derives the interaction energy and solvation free energy for the receptor, ligand and complex respectively. The energy of each molecular is composed of several terms, including van der Waals force (VDWAALS), electrostatic energy (EEL), the electrostatic contribution to the solvation free energy (EGB) and nonpolar contribution to the solvation free energy (ESURF). The total binding free energy is given by ΔG along with error values.
# Results and discussion
# Data analysis
According to the naming rules, 112 EGFR mutation types of the 942 NSCLC patients are divided into five groups, including insertion, deletion, duplication, modification and substitution. We counted the number of mutation types as well as the corresponding patients of each mutation type. From, substitution accounts for more than half of EGFR mutation types and the number of patients. Although deletion just takes up 5.36% of all the mutation types, 285 cases belong to this group and they hold 30.25% of all the patients.
Among all the 112 mutation types, several of them take up the majority of the patients, such as L858R, delE746_A750 and delL747_P753insS. We listed the top 10 common mutations among the 942 patients in. Top two of them (L858R and delE746_A750) accounts for more than half of all the patients in total. Specifically, 388 cases experienced a mutation of L858R at exon 21, taking up 41.19%, and 264 patients suffered from delE746_A750 (deletion of amino acids at exon 19), which accounts for 28.03%.
Moreover, we analyzed the mutations and the number of patients at each exon. From, the number of occurrences of most mutation types at each exon is less than or equal to 3. For example, there are 14 mutation types at exon 18 with the number of occurrences less than or equal to 3 and only 3 mutations with occurrences more than 3. However, those mutations with the number of occurrences more than 3 often occupy more patients than all other mutations with the number of occurrences less than or equal to 3. For instance, 371 patients have 10 mutation types while just 38 patients have 26 mutation types at exon 19. In addition, exon 19 (32.14%) occupies the most mutation types, and exon 21 (44.48%) accounts for the greatest number of the NSCLC patients.
## Egfr mutant structure prediction
We employed Rosetta to generate the EGFR mutants based on the template structures "2ITY" and "1 M17". As the crystal structures of L858R and G719S are available from PDB, we took the two models as the 3D mutant structures. The procedure of predicting EGFR mutants using Rosetta has been discussed in Point Mutation Modeling and Homology Modeling parts of the Materials and methods Section. After the mutant structures were obtained, we employed sander in Amber to conduct a short 1000 steps of minimization to remove any bad contacts and find out the nearest local minima. Then the refined structures were saved as our predicted EGFR mutants for further analysis.shows the mutation neighborhood of our predicted structures and the WT structure. We employed UCSF Chimerato display these structures.
In the EGFR Mutant Structural Database, computationally predicted structures are provided. These structures were selected according to the full atom energy scoring function. To examine the accuracy of the predicted structures, we made a comparison of the predicted structures and the actual structure of L858R, the most common mutation of the EGFR TK domain. Ten amino acids close to the mutation position were selected in order to see the local differences. Using UCSF Chimera, we aligned the predicted structures (structureE01, structureE02 and struc-tureE03 from http://bcc.ee.cityu.edu.hk/data/EGFR.html) to the actual structure "2ITZ" from PDB.shows the comparison of the actual structure and the structures generated based on Rosetta. The backbones of the predicted structures are consistent with the actual one. In addition, the backbone RMSDs of the three pairs of structures are 0.725 Å, 0.562 Å and 0.559 Å respectively, which confirms the good accuracy of our prediction procedure.
Once the optimized mutant structures were obtained, we aligned these structures to the WT EGFR using Match-Maker in UCSF Chimera. Inhibitors were added to the structures followed by several optimization steps to these mutant-drug complexes using MD simulation. Before MD simulation, the mutant-drug complex should first be solvated in a TIP3P water box (10.0 Å) and the ff99SB force filed was needed in order to describe the molecule interactions. Then a series of refinement operations (minimization, heating, density equilibration and constant pressure equilibration) were conducted with sander in Amber as introduced in the Molecular dynamics (MD) simulation part of the Materials and methods Section.shows the comparison of minimized EGFR mutant-drug complex and WT EGFR-drug complex structures.
## Binding free energy calculation
Binding free energies are calculated based on the trajectories determined during the production MD simulation process. Before running the production MD simulation, we should make sure that the solvated complex has equilibrated. For this, the terms of temperature, density and total energy of the system are examined. Moreover, the protein backbone RMSD is checked in order to see whether the conformational stability has been achieved.shows the verification terms of the mutant delE746_A750 with gefitinib and erlotinib during the equilibration period.
From, the computational processes for the density, temperature and total energy of delE746_A750 with gefitinib and erlotinib are all converged at last, which can be used as evidence for system equilibration. The backbone RMSD is in an acceptable level although it is not converged completely for each system. After the system reaches equilibration, production MD simulation is conducted. Similarly, we still check equilibrium phase space of the system by the density, temperature, total energy and backbone RMSD during the production phase in order to obtain good simulation results. Then, with the trajectories produced in the production MD simulation process, binding free energies are calculated using MM-GBSA in Amber tools. The total binding free energies of WT EGFR and several common mutation types with gefitinib and erlotinib are shown in. Moreover, the standard deviations (SD) and standard error of the mean (SED) are also listed. If the binding free energy of the mutant and drug is low, generally, they are considered binding well with each other, which means the drug can inhibit the activation of EGFR mutant. The total binding free energy is composed of several energy components, including VDWAALS, EEL, EGB and ESURF. shows the distribution of each energy component and the total binding free energies of all the EGFR mutants and the inhibitors. As shown in and B, the total energy and energy components (VDWAALS, EEL and EGB) of the EGFR mutant-gefitinib complexes are distributed in a wider range than those of EGFR mutanterlotinib complexes.
## Comparison of the egfr mutant structural database and other egfr-related databases
Several EGFR-related databases are available publicly, such as the EGFR Mutation Database (http://www.cityofhope.org/egfr-mutation-database), the Catalogue of Somatic Mutations in Cancer (http://cancer.sanger.ac. uk/cancergenome/projects/cosmic/), EGFR Inhibitor Database (http://crdd.osdd.net/raghava/egfrindb/)and the widely used PDB. The EGFR Mutation Database contains the mutant position information as well as the response to inhibitors of the NSCLC patients. The COSMIC stores somatic mutation data of human cancer. These databases just provide the sequence information of the mutations. The EGFR Inhibitor Database contains biological and chemical information of the EGFR inhibitors. PDB provides crystal structures of proteins, nucleic acids, and complex assemblies obtained from experimental methods, such as X-ray or NMR. However, only a few EGFR mutant structures are available because of the high cost of experiments. The EGFR Mutant Structural Database presented in this paper contains 3D structures of 112 kinds of EGFR mutants. Moreover, the binding free energies of the mutant and inhibitors are provided to show the binding affinity. The structural information is very helpful to conduct protein docking, hydrogen bond analysis, and protein-drug complex simulation, which are very important in the studying of drug resistance mechanisms. In our previous work, the molecular mechanisms have been identified from the aspects of geometric properties of mutant structures and the binding free energies with gefitinib and erlotinib. In, with 30 mutant structures generated by Rosetta, we analyzed local surface changes of the binding pocket relative to the wild-type EGFR using alpha shape modeling. Moreover, we conducted a correlation analysis about the geometric properties and the pre-recorded progression-free survival (PFS) in the treatments. Results show that the curvature of the binding pocket surface plays an important role in the prediction of EGFR mutation-induced drug resistance. In, we identified drug resistance mechanisms from the binding free energies with inhibitors (gefitinib and erlotinib) as well as some personal features of 168 patients (belonging to 37 mutation types). Extreme learning machine method was employed to build a classification model and resistant subjects were successfully identified. Overall, the molecular mechanisms of drug resistance are closely related to the mutant structures and the binding affinity with inhibitors. Thus, the EGFR Mutant Structural Database we built here is very useful to other researchers and medical doctors for further studying or clinical guidance.
# Conclusions
In this work, we created an EGFR Mutant Structural Database, composed of computationally predicted 3D structures of the EGFR mutants and the corresponding binding free energies with gefitinib and erlotinib. In our database, 112 kinds of mutants were collected from 942 NSCLC patients. We categorized the mutants into five groups (insertion, deletion, duplication, modification and substitution), and substitution accounts for 61.61% of the EGFR mutation types and 54.14% of all the patients. As the most common mutation type, L858R covers 388 or 41.19% of all the patients. In addition, we analyzed the mutations at each exon. It shows that exon 19 (32.14%) possesses the most mutation types and exon 21 (44.48%) occupies the largest number of patients. With the mutant protein sequences and WT EGFR crystal structure, we predicted the EGFR mutation structures with Rosetta and optimized the structures using Amber. Finally, we calculated the binding free energies of EGFR mutants and the inhibitors (gefitinib and erlotinib). Our work provides a database of the EGFR mutant structures and their corresponding binding free energy with inhibitors. These resources can be used for further researches and clinical guidance, such as analyzing drug resistance of the EGFR mutants, which is a major problem during the treatment of NSCLC patients. The database is freely available at http://bcc.ee.cityu.edu.hk/data/EGFR.html. |
Reciprocity between Regulatory T Cells and Th17 Cells: Relevance to Polarized Immunity in Leprosy
T cell defect is a common feature in lepromatous or borderline lepromatous leprosy (LL/BL) patients in contrast to tuberculoid or borderline tuberculoid type (TT/BT) patients. Tuberculoid leprosy is characterized by strong Th1-type cell response with localized lesions whereas lepromatous leprosy is hallmarked by its selective Mycobacterium leprae specific T cell anergy leading to disseminated and progressive disease. FoxP3+ Regulatory T cells (Treg) which are essential for maintaining peripheral tolerance, preventing autoimmune diseases and limiting chronic inflammatory diseases also dampen proinflammatory T cells that include T helper 17 (Th17) cells. This study is aimed at evaluating the role of Treg cells in influencing other effector T cells and its relationship with the cytokine polarized state in leprosy patients. Peripheral blood mononuclear cells from of BT/TT (n = 15) and BL/LL (n = 15) patients were stimulated with M. leprae antigen (WCL) in presence of golgi transport inhibitor monensin for FACS based intracellular cytokine estimation. The frequency of Treg cells showed >5-fold increase in BL/LL in comparison to BT/TT and healthy contacts. These cells produced suppressive cytokine, IL-10 in BL/LL as opposed to BT/TT (p = 0.0200) indicating their suppressive function. The frequency of Th17 cells (CD4, CD45RO, IL-17) was, however, higher in BT/TT. Significant negative correlation (r = -0.68, P = 0.03) was also found between IL-10 of Treg cells and IL-17+ T cells in BL/LL. Blocking IL-10/TGF-β restored the IL-17+ T cells in BL/LL patients. Simultaneously, presence of Th17 related cytokines (TGFβ, IL-6, IL-17 and IL-23) decreased the number of FoxP3+ Treg cells concomitantly increasing IL-17 producing CD4+ cells in lepromatous leprosy. Higher frequency of Programmed Death-1/PD-1+ Treg cells and its ligand, PDL-1 in antigen presenting cells (APCs) was found in BL/LL patients. Inhibition of this pathway led to rescue of IFN-γ and IL-17 producing T cells. Results indicate that Treg cells are largely responsible for the kind of immunosuppression observed in BL/LL patients. This study also proves that Treg cells are profoundly affected by the cytokine milieu and this property may be utilized for benefit of the host.PLOS Neglected Tropical Diseases |Polarized T cell response (Th1/Th2 biased) to Mycobacterium leprae (M. leprae) is believed to be a critical element in the pathogenesis of leprosy and its varied clinical manifestations. However, immune response at the pathologic sites of leprosy is an extremely complex process, particularly in the light of recently evidenced heterogeneity of T cell subsets. FoxP3 positive regulatory T cells (Treg) are one of the most potent hierarchic cell types suppressing the effector T cell function with eventual regulation of immune response elicited by the host during intracellular infections. This study shows the recovery of the cell mediated response by CD4+ T cells by inhibiting the suppressive cytokines, IL-10 and TGF-β and also by blocking of the Programmed Death-1 pathway in cells isolated from lepromatous leprosy patients. Reversal of IL-17 immune response was also achieved by modulating the cytokine milieu of in vitro cell culture and hence provides us cues to counter the M. leprae unresponsiveness in leprosy patients.
# Introduction
Leprosy is a disease of immunological spectrum tightly correlating with the extent of pathology and clinical manifestation [bib_ref] Classification of leprosy according to immunity. A five-group system, Ridley [/bib_ref]. It is well known that T cell defect is a distinctive feature in lepromatous leprosy (LL) in contrast to that of tuberculoid leprosy (TT) patients. In between these clinical entities lie borderline tuberculoid (BT), borderline lepromatous (BL) and borderline borderline (BB) all displaying symptoms in between the two polarized forms [bib_ref] Histological classification and the immunological spectrum of leprosy, Ridley [/bib_ref]. Selective T cell unresponsiveness to the antigens of M. leprae occurs among LL patients, while responsiveness to several other antigens remains intact, a phenomenon known as "split anergy" [bib_ref] Sustained T-cell reactivity to Mycobacterium tuberculosis specific antigens in 'split-anergic' leprosy, Kaleab [/bib_ref]. BT/ TT patients with strong T cell reactivity against M. leprae is associated with biased production of IFN-γ dominant immune response, while BL/LL patients, so called anergic and disseminated form of the disease demonstrates T cell response skewed towards IL-4 and/or IL-10 dominant cytokine production [bib_ref] Th1-Th2 paradigm: insights from leprosy, Modlin [/bib_ref]. Polarized immunity against M. leprae is a critical element in the pathogenesis of leprosy and plays an important role in the varied clinical manifestations of leprosy [bib_ref] Immunopathology of leprosy-current status, Sengupta [/bib_ref]. Biased cytokine production has also been documented at the lesional levels of both TT as well as LL forms of leprosy [bib_ref] Learning from Leprosy: Insights into Contemporary Immunology from an Ancient Disease, Modlin [/bib_ref]. However, generation of Th1/Th2-like effector cells alone cannot fully explain the polarized state of immunity. Other subsets of T cells have been identified which play important role in determining host immunity [bib_ref] Suppressor T lymphocytes from lepromatous leprosy skin lesions, Modlin [/bib_ref] [bib_ref] Cloned suppressor T cells from a lepromatous leprosy patient suppress Mycobacterium leprae..., Ottenhoff [/bib_ref].
Lately, FoxP3 positive regulatory T cells (Tregs) have been characterized as one of the most potent hierarchic cell type suppressing effector T cell function with eventual regulation of immune response elicited by the host during intracellular infections such as tuberculosis [bib_ref] FoxP3+ regulatory T cells suppress effector T-cell function at pathologic site in..., Sharma [/bib_ref] and leishmaniasis [bib_ref] Role for CD4(+) CD25(+) regulatory T cells in reactivation of persistent leishmaniasis..., Mendez [/bib_ref] [bib_ref] Regulatory T cells Suppress T cell Activation at the Pathologic Site of..., Rai [/bib_ref]. The CD4+CD25+ natural regulatory Treg cells expressing the transcription factor forkhead box P3 (FoxP3) is the best characterized suppressive T-cell subset [bib_ref] IL-10-producing and naturally occurring CD4+ Tregs: limiting collateral damage, O'garra [/bib_ref]. These cells are critical for the maintenance of self-tolerance and play an important role in a wide range of clinical conditions such as autoimmune diseases, transplantation rejection reactions, cancer, as well as infectious diseases [bib_ref] Natural regulatory T cells in infectious disease, Belkaid [/bib_ref] [bib_ref] Regulatory T cells and infection: dangerous necessity, Belkaid [/bib_ref]. Mediators of Treg-cell induced suppression include the inhibitory cytokines, IL-10 and TGF-β [bib_ref] Potential role of interleukin-10-secreting regulatory T cells in allergy and asthma, Hawrylowicz [/bib_ref] [bib_ref] Interleukin-10 in the regulation of T cell-induced colitis, Annacker [/bib_ref]. Over representation of Treg cells in the periphery and particularly at the pathologic sites of infection has been shown to be critical in determining local immunity, thus dictating the outcome of the disease among patients suffering from various forms of tuberculosis [bib_ref] FoxP3+ regulatory T cells suppress effector T-cell function at pathologic site in..., Sharma [/bib_ref]. Recently, it was revealed that FoxP3 + inducible Tregs producing TGF-β may down regulate T cell responses leading to the characteristic antigen specific anergy associated with lepromatous leprosy. However, the role of Treg cells in leprosy in association with other subsets needs to be investigated. Treg cells induced by the Programmed Death-1 (PD-1) pathway that assists in maintaining immune homeostasis and prevent autoimmune attack [bib_ref] The PD-1 pathway in tolerance and autoimmunity, Francisco [/bib_ref] may also lead to cellular anergy in lepromatous leprosy. PD-1 is a negative costimulatory molecule which exerts inhibitory effect on T cells by reducing cytokine production and cellular proliferation, with significant effects on IFN-γ, TNF-α and IL-2 production [bib_ref] The PD-1 pathway in tolerance and autoimmunity, Francisco [/bib_ref]. PD-1 may exert its influence on cell differentiation and survival directly through induction of apoptosis [bib_ref] PD-1:PD-L inhibitory pathway affects both CD4+ and CD8+ T cells and is..., Carter [/bib_ref]. The PD-1-PD-L pathway also plays a key role in chronic infections as well as in the suppressive tumor microenvironment [bib_ref] Costimulatory and coinhibitory receptors in anti-tumor immunity, Driessens [/bib_ref] by contributing directly to T-cell exhaustion and lack of immune response [bib_ref] Redefining chronic viral infection, Virgin [/bib_ref]. The importance of the PD-1-PD-L pathway in dampening the T cell responses during chronic infections has propelled the development of strategies to restore host immunity.
Treg cells are well known as inhibitor of various effector T cells such as IFN-γ producing Th1 and recently identified IL-17+ T helper 17 (Th17) cells [bib_ref] CD4+CD25+ regulatory T cells suppress Th17-responses in an experimental colitis model, Ogino [/bib_ref]. Th17 is relatively a new lineage of effector T cells [bib_ref] A brief history of T(H)17, the first major revision in the T(H)1/T(H)2..., Steinman [/bib_ref] and has been shown to be important in immune responses to various infectious and autoimmune diseases [bib_ref] IL-17 and Th17 Cells, Korn [/bib_ref]. Th17 cells mediate their pro-inflammatory function by i) recruiting neutrophils, ii) activating macrophages, and iii) enhancing Th1 effector cells [bib_ref] Induction and effector functions of T(H)17 cells, Bettelli [/bib_ref]. Much of the inflammatory damage previously ascribed to type 1 response is now thought to depend on IL-17 and IL-23 (the cytokine responsible for supporting Th17 response in vivo) [bib_ref] The interleukin-23 axis in intestinal inflammation, Ahern [/bib_ref]. A recent work on leprosy [bib_ref] CD4+ Th17 Cells Discriminate Clinical Types and Constitute a Third Subset of..., Saini [/bib_ref] identified CD4+ Th17 Cells in borderline cases and highlighted their importance in infectious diseases as well. However, little is known about the regulation of this effector T cell subset, especially in relation to Treg cells.
Treg and Th17 cells are therefore two lymphocyte subsets with counterbalancing effects on each other. Altered regulation between these two may contribute to the pathophysiology of infectious diseases by tipping the balance toward suppression of host immunity and subsequent dissemination of disease. Here, we attempted to identify and enumerate these cells in leprosy and the specific factors that impact their cell numbers and function and whether this can be modulated for benefit of the host. Recent findings [bib_ref] Protection of Tregs, suppression of Th1 and Th17 cells, and amelioration of..., Mondal [/bib_ref] especially in autoimmune diseases, suggest that the equilibrium between Treg and Th17 cells might be pharmacologically restored for therapeutic benefit.
# Materials and methods
## Patients and controls
A total of 40 newly diagnosed untreated leprosy patients (23 males, 17 females aged between 19-60 years) from Leprosy Clinic of the Department of Dermatology, All India Institute of Medical Sciences Hospital, New Delhi, India were included in the study [fig_ref] Table 1: Clinical details of 40 newly diagnosed untreated leprosy patients and 10 healthy... [/fig_ref] and classified on the basis of Ridley-Jopling classification [bib_ref] Classification of leprosy according to immunity. A five-group system, Ridley [/bib_ref] [bib_ref] Histological classification and the immunological spectrum of leprosy, Ridley [/bib_ref]. Study group included 20 borderline tuberculoid (BT), 20 lepromatous/borderline lepromatous (LL/BL) patients. Exclusion criteria included patients below 15 years of age, pregnant women, clinical evidence of anemia and other infections such as tuberculosis, HIV and helminthic infestation. PBMCs were investigated for flow cytometric analysis. 10 Healthy Contacts (HCs) were also included in the study.
# Ethics statement
Written informed consent was obtained from all study subjects. The research project was approved by the Institutional Ethics Committee
## Antibodies and reagents
## Pbmc isolation and in vitro cultures
Peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood by Ficol hypaque (Sigma-Aldrich, St. Louis, USA) gradient centrifugation and suspended in RPMI-1640 (Caisson Laboratories, USA) supplemented with L-Glutamine (Sigma, USA), HEPES (Sigma, USA), antibiotics (Biological Industries, Israel) and 10% heat inactivated fetal calf serum (FCS) (Biological Industries, Israel) as per previously published protocol from our laborartory [bib_ref] FoxP3+ regulatory T cells suppress effector T-cell function at pathologic site in..., Sharma [/bib_ref]. PBMCs were cultured (2x10 6 cells/ml) with either PMA+Ionomycin (as positive control) for 6 hours or M. leprae antigen (WCL, 20μg/ml) for 48 hours at 5% CO 2 , 37°C and 1μM Monensin (Sigma, USA) added for last 6 hours. At the end of culture period, cells were stained for intracellular cytokines IL-10, IL-17, IFN-γ and the transcription factor FoxP3 and cell surface markers CD4, CD25, CD45RO, PD-1, PDL-1, CCR4 and CCR6. The cells were finally suspended in staining buffer and acquired on BD FACS Calibur; USA and analysis performed on Flowjo software.
## In vitro blocking experiments
In vitro blocking cell culture experiment was performed in leprosy patients with LL (n = 5) under different conditions, unstimulated vs. stimulated with leprosy antigen only vs. stimulated with antigen and blocked with purified antibodies against IL-10/TGF-β (BD Biosciences) alone or in combination. The experiment was performed in duplicates in 96 well culture plate at 5% CO 2 and 37°C. After an incubation period of 72 hours, the readout was measured in terms of the cytokine IL-17 produced by CD4+ T cells through flow cytometric analysis.
In another experiment, PBMCs isolated from LL patients (n = 4) were incubated with or without blocking abs against PD-1 (5μg/ml, eBioscience), and/or PDL-1 (2μg/ml, eBioscience) to block the interaction between PD-1 and its ligand, in the presence or absence of M. leprae WCL and Brefeldin (GolgiPLUG 10μg/ml, Sigma, USA). Purified mouse IgG1 (final concentration of 10 mg/ml; eBioscience) was used as an isotype control. Cells were cultured in duplicates in 96 well culture plate at 5% CO 2 and 37°C. After an incubation period of 72 hours, cells were examined for the percentage of IFN-γ and IL-17 secreting T cells by flow cytometry. Surface and intracellular staining protocol were performed as mentioned earlier.
Stained cells were then acquired on BD FACS Calibur; USA and analysis performed in Flowjo software. Cell culture with recombinant proteins
For ex vivo cell culture assay using recombinant cytokines, freshly isolated PBMCs from LL patients (n = 5) were cultured for 72 hours in duplicates in 96 well culture plate at 5% CO 2 and 37°C using M. leprae WCL antigen for stimulation. This was done with or without the presence of recombinant cytokines inducing Th17 population such as TGF-β, IL-6, IL-23 and the ones produced by Th17 like IL-17 and IL-22. Cytokines were obtained from eBiosciences and used at concentrations as recommended by the manufacturer in different wells. At the end of culture period, flow cytometry was used to measure the frequency of FoxP3+ CD4+ cells on one hand with simultaneous IL-17 production by CD4+T cells.
# Statistical analysis
FACS analysis was performed with Flowjo software (Tree Star, Oregon, USA). Statistical analysis was performed using GraphPad Prism5 software (La Jolla, CA, USA) applying Student's t test for unpaired samples. Values of p<0.05 were considered significant.
# Results
## Enhanced il10 production and ccr4 expression by treg cells in bl/ll patients
In order to identify the Treg cells in human leprosy, immunophenotyping was done for the markers CD4, CD25 and FoxP3 [fig_ref] Fig 1: T regulatory cells in Leprosy [/fig_ref] on PBMCs derived from leprosy patients (BT/TT vs. BL/LL) (n = 15 in each group) and healthy contacts (HCs, n = 10). Peripheral representation of Treg cells was profoundly increased among BL/LL as compared to BT/TT patients and HCs (p = 0.0002, p<0.0001 respectively) [fig_ref] Fig 1: T regulatory cells in Leprosy [/fig_ref]. Enrichment of Treg cells among BL/LL patients hints towards their important role in the characteristic immune suppression observed in them.
To evaluate the function of enriched Treg cells in BL/LL patients, we measured their intracellular IL-10 production by flow cytometry [fig_ref] Fig 1: T regulatory cells in Leprosy [/fig_ref] , using the whole cell lysate of M. leprae (WCL) for antigen specific stimulation. Percentage of IL-10+ Treg cells (CD4+FoxP3+) was significantly higher among BL/LL patients [fig_ref] Fig 1: T regulatory cells in Leprosy [/fig_ref] compared to that of BT/TT patients (P = 0.0003). The increased frequency of IL-10 producing Treg cells in BL/LL patients indicates their suppressive role in a contact-independent manner via release of IL-10. The surface expression of the chemokine receptor, CCR4 was also significantly high on Treg cells in leprosy patients with BL/LL [fig_ref] Fig 1: T regulatory cells in Leprosy [/fig_ref]. This again highlights that Treg cells are possibly recruited by means of this receptor to the lesional sites in BL/LL patients, wherein they mediate their characteristic suppressive action.
## Elevated il-17 and ccr6 expression on cd4+cd45ro+ cells in bt/tt patients
The Th17 cells are distinct from the classical Th1 and Th2 cell subsets and identified by production of signature cytokine, IL-17. PBMCs isolated from leprosy patients were stimulated with PMA+Ionomycin (as positive control) and M. leprae antigen (WCL) separately. Intracellular staining for IL-17 [fig_ref] Fig 2: Identification of Th17 cells in Leprosy [/fig_ref] on CD4+CD45RO+ T cells showed significantly higher frequency of Th17 cells in BT/TT (P = 0.0005) as compared to BL/LL patients [fig_ref] Fig 2: Identification of Th17 cells in Leprosy [/fig_ref]. This hints towards a possible role of these cells in up-regulation of cell mediated immunity in BT/TT patients leading to containment of disease in the infected host. We also evaluated the expression of CCR6, a well-known chemokine receptor on Th17 cells and observed significant levels of CCR6 surface expression on gated CD4+CD45RO+ cells (P = 0.03) in BT/TT as against BL/ LL patients [fig_ref] Fig 2: Identification of Th17 cells in Leprosy [/fig_ref]. This indicates that Th17 cells plausibly get recruited at the pathologic sites of BT/TT patients and potentiates the Th1 like effector T cells.
## Il-10+ treg cells inversely correlate with th17 cells in bl/ll patients
Both Treg and Th17 cells are known to have important implications in determining the host immunity and subsequent disease manifestation. Hence, we attempted to correlate these two cell types, in both BT/TT and BL/LL patients. In case of BL/LL, there existed a significant inverse correlation (r = -0.68, P = 0.03) between these two T cell subsets [fig_ref] Fig 3: Inverse Correlation between IL-10 and IL-17 in BL/LL [/fig_ref] , which indicates that high IL-10 production in BL/LL correlates with suppressed state of inflammatory immune response as evidenced by lesser number of IL-17+ Th17 cells. However, in BT/TT patients no correlation was found between these two cytokine producing T cell subsets [fig_ref] Fig 3: Inverse Correlation between IL-10 and IL-17 in BL/LL [/fig_ref]. Our results indicate that these two cell types may play critical roles in shaping the host immune response and relative frequencies of IL-10 vs. IL-17 producing T cells can determine the disease outcome. This could also be exploited as a sensitive and specific pathologic biomarker for disease susceptibility in human hosts.
## Il-10 & tgf-β mediated suppression of th17 cells in bl/ll patients
Significant association of the suppressive cytokine IL-10 produced by CD4+FoxP3+ cells with IL-17 by CD4+CD45RO+ T cells in BL/LL patients prompted us to investigate the effect of blocking the cytokines produced by Treg cells, mainly IL-10 and TGF-β on Th17 cells. In BL/ LL patients (n = 5), in vitro blockade of IL-10 and TGF-β alone resulted in elevated frequency of Th17 cells, while dual blockade led to significant increase in IL-17+ T cells [fig_ref] Fig 4: IL-10/TGF-β suppresses Th17 in BL/LL [/fig_ref]. This suggests restoration of the effector immune response in terms of IL-17 production by manipulating the Treg cell derived suppressive cytokines, IL-10 and/or TGF-β in BL/LL patients.
## Th17 cytokines reduces the frequency of foxp3+ treg cells in bl/ll patients
Differentiation of naive T cells towards a Th17 phenotype is induced by several cytokines including TGF-β, IL-1β, IL-6, IL-21, and IL-23 in mice and humans. Hence, we performed ex vivo cell culture assays using PBMCs from BL/LL patients (n = 5) in the presence or absence of Th17 inducing cytokines, TGF-β, IL-6, IL-23 and cytokines produced by Th17, IL-17 and IL-22 [fig_ref] Fig 5: Effect of Th17 cytokines on FoxP3+Treg cells in BL/LL [/fig_ref]. Results showed marked reduction in the frequency of FoxP3+ cells [fig_ref] Fig 5: Effect of Th17 cytokines on FoxP3+Treg cells in BL/LL [/fig_ref] in the presence of Th17 inducing cytokines (TGF-β, IL-6, IL-23) and also in presence of cytokines produced by Th17 (IL-17 and IL-22). Simultaneously, there was an increase in IL-17 producing CD4+ cells in BL/LL patients in these same experimental conditions [fig_ref] Fig 5: Effect of Th17 cytokines on FoxP3+Treg cells in BL/LL [/fig_ref] , hinting that Treg cells are affected by the cytokine milieu they are exposed to. Our result suggests that by means of these cytokines associated with Th17 cells non Treg cells) of BL/LL patients expressed PD-1 on their surface compared to that of BT/TT patients and healthy contacts [fig_ref] Fig 6: PD-1/PDL-1 Interaction Suppress IFN-γ and IL-17 Effector cytokines in BL/LL [/fig_ref]. However, percent PD1+ cells were more in Treg cells relative to FoxP3-T cells (non-Treg) in BL/LL cases. Subsequent assessment of its ligand (PDL-1) revealed higher frequency of CD14+ Monocytes and CD19+ B cells expressing the PDL-1 [fig_ref] Fig 6: PD-1/PDL-1 Interaction Suppress IFN-γ and IL-17 Effector cytokines in BL/LL [/fig_ref]. This data indicates towards a possible role of PD-1/PDL-1 interaction in Treg cell mediated suppression of effector T cells function in BL/LL patients.
To validate the role of PD-1 pathway, we performed in vitro experiments using PBMCs from BL/LL patients (n = 4) stimulated with M. leprae antigen in presence or absence of blocking monoclonal antibodies against PD-1/PDL-1 and measured the intracellular inflammatory cytokines IFN-γ and IL-17 production by CD4+T cells. When PD-1 pathway was blocked, significantly higher number of T cells produced both the pro-inflammatory cytokines, IFN-γ and IL-17 indicating that PD-1/PDL-1 interaction inhibits Th1 and Th17 effector cells in BL/LL patients [fig_ref] Fig 6: PD-1/PDL-1 Interaction Suppress IFN-γ and IL-17 Effector cytokines in BL/LL [/fig_ref]. Considering the higher frequency of PD1+ Treg cells in the BL/LL cases, we envisage that PD1 expressed by the Treg cells dampens the T cell function in these patients. However, PD1 expressed on other subsets of cells may also contribute to the suppression in these cases.
# Discussion
Regulatory T (Treg) cells are essential for maintaining peripheral tolerance, preventing autoimmune diseases and limiting chronic inflammatory diseases [bib_ref] Immunologic self-tolerance maintained by activated T cells expressing IL-2 receptor α-chains (CD25), Sakaguchi [/bib_ref] [bib_ref] Immunologic tolerance maintained by CD25+ CD4+ regulatory T cells: their common role..., Sakaguchi [/bib_ref]. However, in case of chronic infections, they also dampen the host immune response against the pathogen(s) [bib_ref] The role of natural regulatory T cells in infection, Sanchez [/bib_ref]. During an infection, immune regulation is the result of the host's immune response to the infection in a bid to maintain or restore a homeostatic environment [bib_ref] The lifestyle of naturally occurring CD4+ CD25+ Foxp3+ regulatory T cells, Shevach [/bib_ref]. This property is exploited by the pathogen for evading host effector response hence promoting its survival [bib_ref] Regulatory T cells: friend or foe in immunity to infection?, Mills [/bib_ref]. During M. leprae infection varied effector T cell response regulates such balance in a way either containing the pathogen or rendering their survival via immune suppression as in BT/TT and BL/LL respectively [bib_ref] T-Cell Regulation in Lepromatous Leprosy, Bobosha [/bib_ref]. Treg cells suppress effector T cells like Th1 and Th17 cells both by contact dependent as well as contact independent manner. Inhibitory cytokines, IL-10 and TGF-β, have been the focus of considerable attention as mediators of Treg cell induced suppression [bib_ref] Potential role of interleukin-10-secreting regulatory T cells in allergy and asthma, Hawrylowicz [/bib_ref] [bib_ref] Interleukin-10 in the regulation of T cell-induced colitis, Annacker [/bib_ref].
Differential trafficking of these regulatory T cells to the diseased sites are thought to be under the influence of tissue chemokine response [bib_ref] Unique Chemotactic Response Profile and Specific Expression of Chemokine Receptors Ccr4 and..., Iellema [/bib_ref] elicited at the site of LL lesions. Recently we demonstrated that tissue chemokine response at the lepromin DTH site and lesions of various forms of leprosy determines the recruitment of effector T cells at the lesional levels in leprosy patients [bib_ref] Induction of lepromin reactivity in cured lepromatous leprosy patients: impaired chemokine response..., Mitra [/bib_ref]. Some of these subsets have been demonstrated to be hierarchic in nature and exert significant influence on the effector T cells thus regulating the immune response at the pathologic site(s). These include the FoxP3 positive regulatory T cells and our findings validate their critical role in dampening the host effector T cell response during intracellular infections such as tuberculosis [bib_ref] FoxP3+ regulatory T cells suppress effector T-cell function at pathologic site in..., Sharma [/bib_ref] and leishmaniasis [bib_ref] Regulatory T cells Suppress T cell Activation at the Pathologic Site of..., Rai [/bib_ref]. A study in leprosy revealed that FoxP3+ Treg cells that produce TGF-β down regulate the T cell responses and results in the characteristic antigen specific anergy observed in LL cases. However, the role of PD1 in Treg mediated immunosuppression in BL/LL still remains to be studied.
In our study, the frequency of Treg cells showed >5-fold increase in case of BL/LL compared to BT/TT and HCs (p = 0.0002, p<0.0001 respectively). This is in corroboration with previous data from leprosy [bib_ref] Increased Expression of Regulatory T Cells and Down-Regulatory Molecules in Lepromatous Leprosy, Palermo [/bib_ref] [bib_ref] Increased Frequency of CD4 and CD8 Regulatory T Cells in Individuals under..., Fernandes [/bib_ref]. These Treg cells also produce significant amounts of IL-10 in BL/LL vs. BT/TT indicating that these cells are suppressive in nature. Further analysis of surface expression of the chemokine receptor, revealed higher frequency of CCR4+ Treg cells in BL/LL cases as compared to BT/TT. Considering the role of CCR4 in Treg cell homing, we propose that it possibly aids in rapid recruitment of Treg cells at BL/LL lesions. Such local enrichment of Treg cells may facilitate dissemination of disease.
For long, Th1 cells were considered to be the major effectors in multiple autoimmune diseases, while Th2 cells were involved in atopy and asthma [bib_ref] TH1 and TH2 cells: different patterns of lymphokine secretion lead to different..., Mosmann [/bib_ref] [bib_ref] For better or worse: common determinants influencing health and disease in parasitic..., Dent [/bib_ref] [bib_ref] The paradigm of Th1 and Th2 cytokines: its relevance to autoimmunity and..., Singh [/bib_ref]. Recently Th17 cells have been implicated in inflammatory diseases [bib_ref] Differentiation and function of Th17 T cells, Stockinger [/bib_ref]. These cells produce IL-17A (also referred to as IL-17), IL-17F, and IL-22, cytokines involved in neutrophilia, tissue remodeling and repair, and production of antimicrobial proteins [bib_ref] IL-17 and Th17 Cells, Korn [/bib_ref] [bib_ref] Induction and effector functions of T(H)17 cells, Bettelli [/bib_ref]. In our study on leprosy, significantly higher frequency of IL-17+ was found on CD4+CD45RO+ cells in BT/TT (P = 0.0005) as opposed to BL/LL patients. This is indicative of their possible role in inflammatory T cell response observed in BT/TT patients. These cells are also known to express the chemokine receptor CCR6 and we found its expression of Th17 cells (P = 0.03) in BT/TT as against BL/LL. This indicates possible homing of IL-17+ effector T cells in the BT/TT lesions through CCR6. Thus, chemokine receptor mediated reciprocal recruitment of Treg vs. Th17 cells may determine the lesional level immune response and clinical manifestation (localized vs. disseminated in BT/TT vs. BL/LL respectively).
Imbalance between Th17 and Treg cell function may be critical in the immunopathogenesis of many disease states [bib_ref] The Treg/Th17 Cell Balance: A New Paradigm for Autoimmunity, Eisenstein [/bib_ref]. We therefore, correlated the IL-10+ Treg cells with IL-17+ helper T cells and found a significant inverse correlation in BL/LL patients. This indicates that tilting of the balance towards Treg cells may be responsible for the cellular anergy observed in these patients. However, such correlation among BT/TT patients was not significant. This encouraged us to block the Treg cell secreted cytokines, IL-10/TGF-β in vitro in the PBMC culture of BL/LL patients and see their impact on the effector immune response in terms of IL-17. Blocking of each of these cytokines alone or in combination resulted in rescue IL-17+ T cells in BL/ LL. This indicates that by negating the influence of suppressive cytokines we can successfully restore the beneficial effector T cells response in lepromatous leprosy cases.
We further evaluated the impact of Th17 associated cytokines, both Th17 inducing (TGF-β, IL-6, IL-23) and those secreted by Th17 cells (IL-17 and IL-22) on Treg cells. Presence of inducing (TGF-β, IL-6, IL-23) and secreted (IL-17 and IL-22) cytokines markedly reduced the number of Treg cells determined by reduced FoxP3 positivity on CD4+CD25+ cells with simultaneous increase of IL-17+ CD4 T cells. This reduction was however, not seen in cells stimulated with TGF-β alone. Lesser frequency of FoxP3+ Treg cells may also be due to possible apoptosis of these cells in a relatively hostile microenvironment. This indicates that antigen-specific Treg cells are largely dependent on the cytokine milieu they are exposed to. Tipping the balance of Treg cells towards Th17 cell phenotype via alteration of cytokine milieu can result in restoration of M. leprae specific effector T cell response and may therefore constitute possible therapeutic strategy for reversal of immune unresponsiveness in lepromatous leprosy.
T cell responses during parasitic infections are tightly controlled by co-stimulatory or coinhibitory molecules [bib_ref] PD-1 expression on HIV-specific T cells is associated with T-cell exhaustion and..., Day [/bib_ref] [bib_ref] PD-1 expression in acute hepatitis C virus (HCV) infection is associated with..., Urbani [/bib_ref] [bib_ref] Expression of B7-H1 on gastric epithelial cells: its potential role in regulating..., Das [/bib_ref]. PD-1-PDL-1 interaction inhibits the effector T cell functions such as, proliferation, cytokine production and survival, thus balancing the tolerance, autoimmunity, infection, and immunopathology [bib_ref] PD-1 regulates self-reactive CD8+ T cell responses to antigen in lymph nodes..., Keir [/bib_ref] [bib_ref] B7-H1, a third member of the B7 family, co-stimulates T-cell proliferation and..., Dong [/bib_ref]. A recent study showed that in vitro blockade of PD-1 signaling with the specific antibody enhanced IFN-γ production by T cells of TB patients on stimulation with M. tuberculosis antigen [bib_ref] Programmed Death (PD)-1:PD-Ligand 1/PD-Ligand 2 Pathway Inhibits T Cell Effector Functions during..., Jurado [/bib_ref]. These findings raise the possibility PBMCs isolated from peripheral blood were blocked with α-PD-1 or with α-PDL-1 monoclonal blocking antibodies separately or in combination and kept in culture for 72 hours with or without M. leprae WCL antigen. From cultured cells, T cells were analyzed for effector cytokine production by flow cytometry. Total number of cells analyzed by flow cytometry were 500,000. Each bar represents a single category. Mean with SEM are shown in each bar while P value< 0.05 was considered to be significant. Data analysis was performed with flowjo software. Statistical analysis was done using Student's t test for unpaired samples.
doi:10.1371/journal.pntd.0004338.g006
Regulatory T Cells and Th17 Cells in Leprosy that the antigen-specific T-cell response is impaired by some inhibitory mechanism, thereby allowing mycobacterial persistence which was recently demonstrated in our laboratory [bib_ref] Inhibiting PD-1 pathway rescues M. tuberculosis specific IFN-γ producing T cells from..., Singh [/bib_ref]. When we analyzed for the surface expression of PD-1 on Treg and non Treg cells, we found that PD-1 is significantly elevated in BL/LL as compared to BT/TT patients. Even the ligand PDL-1 was higher on CD14+ monocytes and CD19+ B cells suggesting their role in host immune impairment in leprosy. This may be one of the contact-dependent mechanisms utilized by Treg cells for immune suppression. The efficacy of this pathway was firmly established after blockade of PD1-PDL1 interaction increased the M. leprae specific IL-17 and IFN-γ producing T cells in BL/LL patients. Hence, inhibition of this pathway may be an efficient means to revive the host immune response.
Taken together, our study highlights the importance of the equilibrium between Treg and Th17 cells in determining host immunity. Providing a balanced level of function for these two important cell subsets is the key to achieving an appropriate level of parasite control without inducing immunopathology. This would be a major goal in the management of this still-challenging infectious disease.
[fig] Fig 1: T regulatory cells in Leprosy. A) Representative FACS plot showing enumeration of Treg cell frequency (CD4+, CD25+ and FoxP3+) in BT/TT vs. BL/LL patient B) Total frequency of circulating Treg (CD4+, CD25+ and FoxP3+) cells in PBMCs isolated from peripheral blood of BT/TT vs. BL/LL patients (n = 20) vs. Healthy Contacts (n = 10). C) Representative flow cytometric analysis showing IL-10 production following in vitro stimulation with M. leprae (WCL) vs. unstimulated for 48 h in a Leprosy patient with LL. D) Scatter Plots are showing total IL-10 production in gated CD4+ FoxP3+ cells in BT/TT vs. BL/LL (n = 15) vs. Healthy Contacts (n = 10). E) CCR4 surface expression on gated CD4+ FoxP3+ cells in different groups of Leprosy patients (BT/TT vs. BL/LL), (n = 10). Each dot represents a single individual. Median values are shown in each set while P value< 0.05 was considered to be significant. Total number of cells analyzed by flow cytometry were 500,000. Data analysis was performed with flowjo software. Statistical analysis was done using Student's t test for unpaired samples. doi:10.1371/journal.pntd.0004338.g001 [/fig]
[fig] Fig 2: Identification of Th17 cells in Leprosy. A) Representative flow cytometric analysis showing IL-17 production on gated CD4+ CD45RO+ memory T cells following in vitro stimulation with M. leprae (WCL) vs. unstimulated for 48 h in a Leprosy patient with BT. 6 h PMA stimulation was used as positive control. B) Scatter plots are showing cumulative IL-17 production in gated CD4+ CD45RO+ cells in different groups of Leprosy patients (BT/TT vs. BL/LL) and Healthy Contacts (n = 10). Each dot represents a single individual of BT/TT (n = 15) and BL/LL (n = 15) patient. C) Scatter graphs are showing CCR6 surface expression on gated CD4+CD45RO+ cells in different groups of Leprosy patients (BT/TT vs. BL/LL) (n = 10). Median values are shown in each set while P value< 0.05 was considered to be significant. Total number of cells analyzed by flow cytometry were 500,000. Data analysis was performed with flowjo software. Statistical analysis was done using Student's t test for unpaired samples.doi:10.1371/journal.pntd.0004338.g002 [/fig]
[fig] Fig 3: Inverse Correlation between IL-10 and IL-17 in BL/LL. A) Treg derived high IL-10 negatively correlates with low IL-17+ CD4 T cells in BL/LL (n = 13) signifying polarized immunity in leprosy. B) No correlation was found in BT/TT (n = 13). Correlation was done using Spearman rank correlation coefficient; r value was equal to -0.65 while p value was significant at 0.01 in case of BL/LL. doi:10.1371/journal.pntd.0004338.g003 [/fig]
[fig] Fig 4: IL-10/TGF-β suppresses Th17 in BL/LL. A) Flow cytometric analysis showing the increased production of IL-17 on total gated CD4+ T lymphocytes in a leprosy patient with LL under different conditions, Unstimulated vs. Stimulated with leprosy antigen only vs. Stimulated with antigen and blocked with antibodies against IL-10/TGF-β alone or in combination. B) Cumulative data of BL/LL (n = 5) patients are shown in the bar graphs with the same conditions as above. Readout was taken in terms of IL-17 by CD4+ T cells at the end of 72 h. Mean with SEM are shown in each bar while P value< 0.05 was considered to be significant. Total number of cells analyzed by flow cytometry were 500,000. Data analysis was performed with flowjo software. Statistical analysis was done using Student's t test for unpaired samples. doi:10.1371/journal.pntd.0004338.g004 Regulatory T Cells and Th17 Cells in Leprosy PLOS Neglected Tropical Diseases | DOI:10.1371/journal.pntd.0004338 January 11, 2016 [/fig]
[fig] Fig 5: Effect of Th17 cytokines on FoxP3+Treg cells in BL/LL. A) PBMCs were cultured for 72 h in different conditions with or without the presence of cytokines inducing Th17 population such as TGF-β, IL-6, IL-23 and those secreted by Th17, IL-17 and IL-22 in different wells and readout was measured in terms of FoxP3+ on CD4+CD25+ cells on one hand and simultaneous IL-17 production by CD4+T cells. B) Bar graphs show the cumulative data in BL/LL (n = 5) of FoxP3+ expression on CD4+ CD25+ cells as against C) IL-17 production by CD4+T cells under the same conditions as mentioned above. Mean with SEM are shown in each bar while P value< 0.05 was considered to be significant. Total number of cells analyzed by flow cytometry were 500,000. Data analysis was performed with flowjo software. Statistical analysis was done using Student's t test for unpaired samples.doi:10.1371/journal.pntd.0004338.g005 [/fig]
[fig] Fig 6: PD-1/PDL-1 Interaction Suppress IFN-γ and IL-17 Effector cytokines in BL/LL. A) Percentage frequency of PD-1 expression on CD4+FoxP3 + Treg cells vs. B) non Treg cells, (CD4+FoxP3-) in BT/TT vs. BL/LL (n = 10) patients and healthy contacts (n = 10). Percentage frequency of PDL-1 expression on APCs, C) on CD14+ Monocytes D) on CD19+ B cells in BT/TT vs. BL/LL (n = 10) patients and healthy contacts (n = 10) in PBMCs isolated from peripheral blood. Each bar represents a single category. Error bars are shown in each set while P value< 0.05 was considered to be significant. Bar graph representation showing the increased production of E) IFN-γ and F) IL-17 on total gated CD4+ T lymphocytes in BL/LL patients of leprosy (n = 4). [/fig]
[table] Table 1: Clinical details of 40 newly diagnosed untreated leprosy patients and 10 healthy contact subjects. Patients were typed on the basis of Ridley Jopling classification [1], BI; Bacillary Index (mean of six lesional sites). M; male, F; female. BT: Borderline Tuberculoid, LL: Lepromatous Leprosy, HC: Healthy house hold contacts with long exposure to leprosy patients. [/table]
|
Tumors and Cytomegalovirus: An Intimate Interplay
# Introduction
Human cytomegalovirus (HCMV) or human herpesvirus 5 (HHV-5), one of the eight human herpesviruses, can establish lifelong latency within its corresponding host, as well as possessing reactivation potential. An exquisite equilibrium between the immunocompetent host and the virus itself allows viral presence for a lifetime parallel to improved viral fitness over time. Besides birth defects and congenital infections in the adult host, HCMV can cause symptoms due to its reactivation in immunosuppressed patients (organ transplant recipients) and immunocompromised patients (septic patients, elderly, HIV-infected patients, etc.) and favor age-related diseases such as atherosclerosis, immunosenescence, and malignancies [bib_ref] Immune Landscape of CMV Infection in Cancer Patients: From "Canonical" Diseases toward..., El Baba [/bib_ref]. Recent investigations have reported the presence of HCMV proteins and DNA in tumoral tissues of glioblastoma, neuroblastoma, colon cancer, breast cancer, ovarian cancer, and prostate adenocarcinoma [bib_ref] Potential Impact of Human Cytomegalovirus Infection on Immunity to Ovarian Tumours and..., Cox [/bib_ref] [bib_ref] Role of Tumor Cell Immune Escape Mechanisms in Cytomegalovirus-Mediated Oncomodulation, Cinatl [/bib_ref] [bib_ref] The Emerging Role of Human Cytomegalovirus Infection in Human Carcinogenesis: A Review..., Söderberg-Nauclér [/bib_ref] [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref] [bib_ref] Macrophages and Breast Cancer, Pasquereau [/bib_ref]. Although some investigators fail to detect HCMV proteins and/or DNA in tumoral tissue, this might be linked to some technical limitations. Thus, the detection of HCMV proteins in tumor tissues is not just a random rare finding. Optimized PCR and immunohistochemistry protocols detect HCMV proteins and DNA in a majority of tissue specimens from a variety of tumors [bib_ref] Detection of Human Cytomegalovirus Proteins in Paraffin-Embedded Breast Cancer Tissue Specimens-A Novel, Touma [/bib_ref] [bib_ref] Söderberg-Nauclér, C. Presence of the Human Cytomegalovirus in Glioblastomas-A Systematic Review, Peredo-Harvey [/bib_ref].
So far, seven human oncogenic viruses have been listed: Epstein-Barr virus (EBV), Kaposi sarcoma human virus (KSHV), human papillomavirus (HPV), hepatitis B and C viruses (HBV, HCV), human T-lymphotropic virus-1 (HTLV-1), and the most recently discovered Merkel cell polyomavirus (MCPyV) in January 2008. Although HCMV is not yet included in the oncogenic viruses list, it becomes more and more likely that HCMV is the eighth human oncovirus [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref] [bib_ref] Polyploid Giant Cancer Cells, Herbein [/bib_ref]. Although Koch's postulation for viruses and cancer and Hill's criteria for causation might be useful, they also have a limited understanding of virus-induced oncogenesis. Among the seven oncoviruses recognized so far, we can discriminate between direct tumor viruses (HPV, KSHV, HTLV1, EBV, HBV, and MCPyV) that either encode viral oncoproteins or activate host oncoproteins and indirect oncoviruses such as HCV, which set the stage for neoplasm mainly by chronic inflammation. This subclassification of oncoviruses might be artificial since some oncoviruses such as HBV produce a viral oncoprotein (HBx) parallel to inflammation and/or immune activation. The definitive classification of HCMV as the eighth oncovirus could be greatly facilitated by the discovery of one or several viral oncoproteins (e.g., such as E6 and E7 for HPV).
HCMV contributes to carcinogenesis as an initiator or promoter [bib_ref] Human Cytomegalovirus IE1 and IE2 Proteins Are Mutagenic and Mediate "Hit-and-Run, Shen [/bib_ref] [bib_ref] A Review of the Potential Role of Human Cytomegalovirus (HCMV) Infections in..., Geisler [/bib_ref]. HCMV allows the tumor to escape immune surveillance by encoding viral proteins and inducing various immunosuppressive cellular factors allowing HCMV-induced immune tolerance, which favors tumor growth. In return, HCMV will increase its fitness in the immunologically weak environment of the cancerous cells [bib_ref] Cytomegalovirus and Tumors: Two Players for One Goal-Immune Escape, Lepiller [/bib_ref]. Recently, HCMV has been reported to trigger the generation of polyploid giant cancer cells (PGCCs), as already reported for all human oncoviruses discovered so far [bib_ref] Polyploid Giant Cancer Cells, Herbein [/bib_ref] [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref]. PGCCs are keystone cancer cells resulting in stemness, chemoresistance, metastasis, and relapse in poor-prognosis cancers [bib_ref] Polyploid Giant Cancer Cells: An Emerging New Field of Cancer Biology, Liu [/bib_ref]. We and others have suggested revisiting the hallmarks of cancer to add a new hallmark, namely PGCCs and giant cell cycling [bib_ref] Polyploid Giant Cancer Cells, Herbein [/bib_ref] [bib_ref] Cancer Recurrence and Lethality Are Enabled by Enhanced Survival and Reversible Cell..., Pienta [/bib_ref] [bib_ref] The Life Cycle of Polyploid Giant Cancer Cells and Dormancy in Cancer:..., Liu [/bib_ref].
Most of the oncoviruses, and especially oncoviruses, of the Herpesviridae family, namely EBV and KSHV, alternate lytic and latent phases to favor tumor initiation and spread [bib_ref] Cancer Progression Goes Viral: The Role of Oncoviruses in Aggressiveness of Malignancies, Müller-Coan [/bib_ref]. Similarly, HCMV infection starts with viremia and lytic cycle, followed by latency in healthy individuals. Nevertheless, the homeostatic equilibrium established between HCMV and the host could be broken time by time by immunosuppressive conditions, during acute infections (bacterial and viral), transplantation, and even in physiological immune suppression such as in pregnancy. Immune imbalance in such immunocompromised patients can result in unhindered viral replication, followed by the reactivation of the latent virus in an inflammatory context [bib_ref] Human Cytomegalovirus Activates Inflammatory Cytokine Responses via CD14 and Toll-Like Receptor 2, Compton [/bib_ref] , which will result in subsequent increased morbidity, but also, when resolved could favor HCMV fitness [bib_ref] Human Cytomegalovirus Immunity and Immune Evasion, Jackson [/bib_ref] [bib_ref] Epigenetic Regulation of Human Cytomegalovirus Latency: An Update, Kumar [/bib_ref]. In fact, during a lifetime period, HCMV, through these successive periods of reactivation and latency, will genetically and phenotypically evolve to adapt to a moving immune landscape [bib_ref] Extensive Genome-Wide Variability of Human Cytomegalovirus in Congenitally Infected Infants, Renzette [/bib_ref] [bib_ref] Human Cytomegalovirus Multiple-Strain Infections and Viral Population Diversity in Haematopoietic Stem Cell..., Dhingra [/bib_ref] [bib_ref] Human Cytomegalovirus Genomes Sequenced Directly from Clinical Material: Variation, Multiple-Strain Infection, Recombination,..., Suárez [/bib_ref]. Various techniques, including restriction fragment length polymorphism (RLFP) analysis [bib_ref] Molecular Epidemiology of Cytomegalovirus Infections in Women and Their Infants, Huang [/bib_ref] , targeted amplicon sequencing [bib_ref] Cytomegalovirus and Epstein-Barr Virus Subtypes-the Search for Clinical Significance, Puchhammer-Stöckl [/bib_ref] [bib_ref] Mixed Cytomegalovirus Glycoprotein B Genotypes in Immunocompromised Patients, Coaquette [/bib_ref] [bib_ref] Analysis of Mixed Infections by Multiple Genotypes of Human Cytomegalovirus in Immunocompromised..., Sowmya [/bib_ref] [bib_ref] Donor Cytomegalovirus Transmission Patterns in Solid Organ Transplant Recipients with Primary Infection, Hasing [/bib_ref] , and whole-genome sequencing [bib_ref] Human Cytomegalovirus Multiple-Strain Infections and Viral Population Diversity in Haematopoietic Stem Cell..., Dhingra [/bib_ref] , have been used to describe HCMV genome variability. With the increasing sequencing depth of next-generation sequencing platforms, the detection of low-frequency variants, i.e., minors, became possible [bib_ref] Puchhammer-Stöckl, E. Deep Sequencing Reveals Highly Complex Dynamics of Human Cytomegalovirus Genotypes..., Görzer [/bib_ref]. Currently, there is mounting evidence that HCMV exists as a heterogeneous collection of genomes with variations in composition and distribution between anatomical compartments [bib_ref] Rapid Intrahost Evolution of Human Cytomegalovirus Is Shaped by Demography and Positive..., Renzette [/bib_ref] [bib_ref] Characterization of Human Cytomegalovirus Genome Diversity in Immunocompromised Hosts by Whole-Genome Sequencing..., Hage [/bib_ref] and over time [bib_ref] Human Cytomegalovirus Multiple-Strain Infections and Viral Population Diversity in Haematopoietic Stem Cell..., Dhingra [/bib_ref] [bib_ref] Puchhammer-Stöckl, E. Deep Sequencing Reveals Highly Complex Dynamics of Human Cytomegalovirus Genotypes..., Görzer [/bib_ref] [bib_ref] Characterization of Human Cytomegalovirus Genome Diversity in Immunocompromised Hosts by Whole-Genome Sequencing..., Hage [/bib_ref]. Performing high-throughput analysis, HCMV genome diversity is significantly more divergent than all other human herpesviruses and highlights the capacity of the viral genome to adapt to its host environment with high flexibility [bib_ref] High-Throughput Analysis of Human Cytomegalovirus Genome Diversity Highlights the Widespread Occurrence of..., Sijmons [/bib_ref]. In addition to the high diversity of HCMV clinical strains [bib_ref] High-Resolution Human Cytomegalovirus Transcriptome, Gatherer [/bib_ref] , a key element to discriminate between oncogenic and oncomodulatory properties of HCMV strains is the limited potential of HCMV to replicate in already transformed cells [bib_ref] Expression of Oncogenic Alleles Induces Multiple Blocks to Human Cytomegalovirus Infection, Xu [/bib_ref]. Thus, HCMV is unable to productively infect most cancer-derived cell lines due to the fact that oncogenic alleles induce multiple restrictions to HCMV replication [bib_ref] Expression of Oncogenic Alleles Induces Multiple Blocks to Human Cytomegalovirus Infection, Xu [/bib_ref]. In contrast, the HCMV-DB and BL clinical strains isolated in our laboratory, which display oncogenic capacities in human epithelial mammary cells (HMECs), can fully replicate in primary epithelial cells with the alternation of both lytic and latent viral cycles and the appearance of CMV-transformed HMECs (CTH cells) up to several months post-infection [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] [bib_ref] Distinct Oncogenic Transcriptomes in Human Mammary Epithelial Cells Infected with Cytomegalovirus, Ahmad [/bib_ref]. Thus, a specific fitness for long-run epithelial cell replication of the HCMV oncogenic strains has to be taken into account to highlight the oncogenic properties of the virus.
During these prolonged periods of latency, numerous viral proteins are at play. For example, in endothelial cells, the UL133-UL138 locus, encoded in the ULb region of the HCMV genome, is essential for the viral late-stage response [bib_ref] An Endothelial Cell-Specific Requirement for the UL133-UL138 Locus of Human Cytomegalovirus for..., Bughio [/bib_ref]. In infected cells, UL135 and UL136 genes promote viral maturation, but most importantly, this locus involves the main molecular switch among latency and reactivation, including the opposing roles of UL135 and UL138 (reviewed in [bib_ref] The Role of the Human Cytomegalovirus UL133-UL138 Gene Locus in Latency and..., Mlera [/bib_ref]. The immune landscape is also essential to shape the outcome of antiviral immunity and is influenced by numerous viral determinants, including HCMV strain, virulence, MHC I downregulation, and other escape strategies elicited by HCMV during the early virus-host interaction [bib_ref] Immune Landscape of CMV Infection in Cancer Patients: From "Canonical" Diseases toward..., El Baba [/bib_ref]. This review highlights the considerable influence of HCMV on the immune landscape, the HCMV oncomodulatory and newly reported oncogenic signals that contribute to cancers, especially of poor prognosis.
## Hcmv and tumor, a "win-win" strategy
HCMV, through its oncomodulatory effect, catalyzes the oncogenic process within the cancerous cells by expressing viral proteins known to possess a pro-oncogenic power such as IE1 and US28, among others, helping tumor cells to evade the immune system, preventing cell death, and favoring cell survival [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref]. In addition, the restricted cellular immune responses against HCMV in these immune-privileged tumor sites will enhance HCMV replication in a context of constrained replication, as reported previously in HCMV infection of transformed cells [bib_ref] Expression of Oncogenic Alleles Induces Multiple Blocks to Human Cytomegalovirus Infection, Xu [/bib_ref]. On the other hand, cancer cells on their own escape immune control favoring epithelial-mesenchymal transition (EMT), metastasis, and relapse. Thus, the combination of the intrinsic cellular machinery and the viral immune escape strategies in cancer cells may offer an environment that enhances limited HCMV replication and boosts cancer cells to evade immune surveillance showing the bidirectional relationship between tumor cells and HCMV [fig_ref] Figure 1: HCMV rewires the immune landscape to favor both HCMV fitness and tumor... [/fig_ref] [bib_ref] Role of Tumor Cell Immune Escape Mechanisms in Cytomegalovirus-Mediated Oncomodulation, Cinatl [/bib_ref] [bib_ref] Cytomegalovirus and Tumors: Two Players for One Goal-Immune Escape, Lepiller [/bib_ref] [bib_ref] Molecular Mechanisms of the Modulatory Effects of HCMV Infection in Tumor Cell..., Cinatl [/bib_ref].
to several months post-infection [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] [bib_ref] Distinct Oncogenic Transcriptomes in Human Mammary Epithelial Cells Infected with Cytomegalovirus, Ahmad [/bib_ref]. Thus, a specific fitness for long-run epithelial cell replication of the HCMV oncogenic strains has to be taken into account to highlight the oncogenic properties of the virus.
During these prolonged periods of latency, numerous viral proteins are at play. For example, in endothelial cells, the UL133-UL138 locus, encoded in the ULb′ region of the HCMV genome, is essential for the viral late-stage response [bib_ref] An Endothelial Cell-Specific Requirement for the UL133-UL138 Locus of Human Cytomegalovirus for..., Bughio [/bib_ref]. In infected cells, UL135 and UL136 genes promote viral maturation, but most importantly, this locus involves the main molecular switch among latency and reactivation, including the opposing roles of UL135 and UL138 (reviewed in [bib_ref] The Role of the Human Cytomegalovirus UL133-UL138 Gene Locus in Latency and..., Mlera [/bib_ref]. The immune landscape is also essential to shape the outcome of antiviral immunity and is influenced by numerous viral determinants, including HCMV strain, virulence, MHC I downregulation, and other escape strategies elicited by HCMV during the early virus-host interaction [bib_ref] Immune Landscape of CMV Infection in Cancer Patients: From "Canonical" Diseases toward..., El Baba [/bib_ref]. This review highlights the considerable influence of HCMV on the immune landscape, the HCMV oncomodulatory and newly reported oncogenic signals that contribute to cancers, especially of poor prognosis.
## Hcmv and tumor, a "win-win" strategy
HCMV, through its oncomodulatory effect, catalyzes the oncogenic process within the cancerous cells by expressing viral proteins known to possess a pro-oncogenic power such as IE1 and US28, among others, helping tumor cells to evade the immune system, preventing cell death, and favoring cell survival [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref]. In addition, the restricted cellular immune responses against HCMV in these immune-privileged tumor sites will enhance HCMV replication in a context of constrained replication, as reported previously in HCMV infection of transformed cells [bib_ref] Expression of Oncogenic Alleles Induces Multiple Blocks to Human Cytomegalovirus Infection, Xu [/bib_ref]. On the other hand, cancer cells on their own escape immune control favoring epithelial-mesenchymal transition (EMT), metastasis, and relapse. Thus, the combination of the intrinsic cellular machinery and the viral immune escape strategies in cancer cells may offer an environment that enhances limited HCMV replication and boosts cancer cells to evade immune surveillance showing the bidirectional relationship between tumor cells and HCMV [fig_ref] Figure 1: HCMV rewires the immune landscape to favor both HCMV fitness and tumor... [/fig_ref] [bib_ref] Role of Tumor Cell Immune Escape Mechanisms in Cytomegalovirus-Mediated Oncomodulation, Cinatl [/bib_ref] [bib_ref] Cytomegalovirus and Tumors: Two Players for One Goal-Immune Escape, Lepiller [/bib_ref] [bib_ref] Molecular Mechanisms of the Modulatory Effects of HCMV Infection in Tumor Cell..., Cinatl [/bib_ref].
## Hcmv decreases viral-specific cd4 + and cd8 + t-cell response, nk activity, which could favor tumor growth
HCMV fitness is a key factor for the virus to modulate the immune landscape. Thus, hindering the MHC class I-restricted antigen presentation is a key mechanism to favor longlasting low replication of the virus [bib_ref] The MHC Class I Antigen Presentation Pathway: Strategies for Viral Immune Evasion, Hewitt [/bib_ref]. The matrix protein, pp65, through its kinase activity, phosphorylates the IE1 protein and specifically inhibits the presentation of IE-derived antigenic peptides to escape immune recognition of the early produced viral proteins [bib_ref] Selective Interference with Class I Major Histocompatibility Complex Presentation of the Major..., Gilbert [/bib_ref]. Knowing that pp65 is delivered directly into the cells during the viral fusion phase, HCMV will immediately escape from immunological surveillance [bib_ref] Viral Immune Evasion: A Masterpiece of Evolution, Vossen [/bib_ref]. In addition, pp65, which triggers the host interferon responses, inhibits NK cell detection or activation [bib_ref] HCMV-Encoded NK Modulators: Lessons from in Vitro and in Vivo Genetic Variation, Patel [/bib_ref] and limits the recognition of CD4 + and CD8 + T cells by preventing MHC class I and II antigen processing and appearance [bib_ref] Viral Immune Evasion: A Masterpiece of Evolution, Vossen [/bib_ref]. Finally, a viral IL-10 (v-IL-10) homolog is produced in infected cells, which will create a Th2 immune landscape and counteract the anti-HCMV CD4 + and CD8 + T-cell responses [bib_ref] Immune Landscape of CMV Infection in Cancer Patients: From "Canonical" Diseases toward..., El Baba [/bib_ref].
To counteract the NK cell response parallel to the limited MHC class I expression observed in the vicinity of HCMV-infected cells, HCMV will display numerous molecular mechanisms. gpUL40 and vIL10, respectively, increase the expression of HLA-E and HLA-G on the cell surface of infected cells [bib_ref] Cutting Edge: The Human Cytomegalovirus UL40 Gene Product Contains a Ligand for..., Ulbrecht [/bib_ref] [bib_ref] Surface Expression of HLA-E, an Inhibitor of Natural Killer Cells, Enhanced by..., Tomasec [/bib_ref] [bib_ref] Potent Immunosuppressive Activities of Cytomegalovirus-Encoded Interleukin-10, Spencer [/bib_ref]. The UL16 protein inhibits natural killer group 2D (NKG2D)-mediated NK cell activation by blocking the binding of NKG2D to UL16-binding proteins (ULBPs), namely ULBP1 and ULBP2, and to the MHC class I chainrelated gene B (MICB gene) [bib_ref] Immunobiology of Human Cytomegalovirus: From Bench to Bedside, Crough [/bib_ref]. US18 and US20 viral proteins increase the destruction of MHC class I polypeptide-related sequence A (MICA), thereby limiting the NK cell from recognizing stress signals elicited by HCMV-infected cells [bib_ref] Two Novel Human Cytomegalovirus NK Cell Evasion Functions Target MICA for Lysosomal..., Fielding [/bib_ref]. To escape NK cell control, HCMV blocks NK cell-activating receptor (NKp30) by pp65, MICB gene expression by UL122-encoded microRNA, and CD155 expression by UL141 [bib_ref] Immunobiology of Human Cytomegalovirus: From Bench to Bedside, Crough [/bib_ref] [bib_ref] Host Immune System Gene Targeting by a Viral MiRNA, Stern-Ginossar [/bib_ref].
In contrast to healthy individuals with high and sustained HCMV-specific CTLs, in chronic HCMV infection, T-cell exhaustion occurs similar to the T-cell exhaustion observed in cancer patients [bib_ref] Programmed Death 1 Signaling on Chronic Myeloid Leukemia-Specific T Cells Results in..., Mumprecht [/bib_ref]. Impaired immunological synapse formation and vesicle trafficking, loss of proliferative capacity, limited cytotoxicity, and decreased cytokine production by effector CD8 + T cells accompany T-cell exhaustion [bib_ref] Molecular Signature of CD8+ T Cell Exhaustion during Chronic Viral Infection, Wherry [/bib_ref]. In addition, in HCMV-infected patients, the appearance of CMV-specific CD4 + CD27 − CD28 − T cells, which function as regulatory T cells (Treg), may explain their deleterious effects in CMV-infected individuals [bib_ref] Understanding the Mechanism of Action of Cytomegalovirus-Induced Regulatory T Cells, Tovar-Salazar [/bib_ref]. Additionally, CD28 − T cells have an inbuilt ability to release inflammatory cytokines and cytotoxic molecules that can damage tissues and amplify inflammatory pathways [bib_ref] The Life (and Death) of CD4+ CD28(Null) T Cells in Inflammatory Diseases, Dumitriu [/bib_ref] [bib_ref] Expansions of Cytotoxic CD4+CD28− T Cells Drive Excess Cardiovascular Mortality in Rheumatoid..., Broadley [/bib_ref]. In addition, CMV infection upregulates Th17 cell activity during acute organ rejection with enhanced proinflammatory cytokine production [bib_ref] NK Cell and Th17 Responses Are Differentially Induced in Murine Cytomegalovirus Infected..., Li [/bib_ref]. Th17 cells exemplify immune adaptation. On the one hand, Th17 cells can have a full effector function protecting the host from pathogens with the associated risk of developing immune-mediated inflammatory diseases with enhanced proinflammatory cytokine production [bib_ref] Interleukin-17 in Host Defence against Bacterial, Mycobacterial and Fungal Pathogens, Curtis [/bib_ref]. Th17 cells can also acquire an anti-inflammatory fate characterized by the secretion of the anti-inflammatory cytokine IL-10 [bib_ref] The Induction and Function of the Anti-Inflammatory Fate of TH17 Cells, Xu [/bib_ref]. Altogether, the increase in Treg, CD28 − T cells, and Th17 cells in chronic HCMV infection could favor tumor relapse and spread in cancer patients parallel to chronic inflammation. Altogether, the impairment of HCMV-specific CD4 + and CD8 + T cells and NK cells function, parallel to the enhancement of the activity of Treg, CD28 − T cells, and Th17 cells could favor both viral spread and tumor development.
## Hcmv favors tumor cell survival
To favor cell survival and promote its own fitness, HCMV blocks the apoptotic machinery within infected cells. Thus, the expression of B-cell lymphoma 2 (Bcl-2) and Fas-associated death-domain-like IL-1β-converting enzyme-inhibitory proteins (FLIP) is upregulated in HCMV-infected cells, especially through IE2 [bib_ref] The Immediate Early 2 Protein of Human Cytomegalovirus (HCMV) Mediates the Apoptotic..., Chiou [/bib_ref] [bib_ref] Protooncogene Bcl-2 Gene Transfer Abrogates Fas/APO-1 Antibody-Mediated Apoptosis of Human Malignant Glioma..., Weller [/bib_ref]. pUL36 blocks the activity of procaspase 8 to the death-inducing signaling complex (DISC). pUL37 limits the action of proapoptotic Bcl-2 members, namely Bcl-2-associated X Protein (Bax) and Bcl-2 homologous antagonist/killer (Bak) [bib_ref] Cytomegalovirus and Tumors: Two Players for One Goal-Immune Escape, Lepiller [/bib_ref] [bib_ref] Viral Immune Evasion: A Masterpiece of Evolution, Vossen [/bib_ref]. Furthermore, HCMV has developed UL36 and UL37 proteins, which enhance the survival of infected cells, thus stimulating viral dissemination within the host [bib_ref] Immune Landscape of CMV Infection in Cancer Patients: From "Canonical" Diseases toward..., El Baba [/bib_ref]. The upregulation of the host-encoded complement regulatory proteins (CRPs) [bib_ref] Altered Expression of Host-Encoded Complement Regulators on Human Cytomegalovirus-Infected Cells, Spiller [/bib_ref] and the integration of host cell-derived CRPs, CD55, and CD59 in virions curtail the complement attack and favor HCMV spread [bib_ref] Cytomegalovirus and Tumors: Two Players for One Goal-Immune Escape, Lepiller [/bib_ref].
## Hcmv buffers the inflammatory landscape and favors a th2/m2 landscape
In addition, to counteract the cellular immune defense mediated through HCMVspecific CTLs and NK cells, HCMV constrains the inflammatory environment present in the vicinity of infected cells. Thus, upon viral entry, HCMV triggers the release of type I interferon (IFN), upregulates CD80 and CD86, and activates the NF-κB pathway lead- ing to the production of inflammatory cytokines interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α) through the interaction of gB and gH with the toll-like receptor 2 (TLR2) [bib_ref] Human Cytomegalovirus Activates Inflammatory Cytokine Responses via CD14 and Toll-Like Receptor 2, Compton [/bib_ref] [bib_ref] Human Cytomegalovirus Immunity and Immune Evasion, Jackson [/bib_ref] [bib_ref] Human Cytomegalovirus Envelope Glycoproteins B and H Are Necessary for TLR2 Activation..., Boehme [/bib_ref]. In addition, UL144, a viral homolog of the TNF receptor and an IL-8-like chemokine (viral CXC-1), which favors the chemotaxis of peripheral blood neutrophils, will further participate in the proinflammatory microenvironment [bib_ref] The UL144 Gene Product of Human Cytomegalovirus Activates NFkappaB via a TRAF6-Dependent..., Poole [/bib_ref]. Such a proinflammatory microenvironment, although it might stimulate viral clearance through the activation of phagocytic cells [bib_ref] The Immune Response to Human CMV, Rosa [/bib_ref] , could also favor cellular transformation since the activation of the IL6/JAK/STAT3 axis could favor the appearance and the development of hepatocellular carcinoma (HCC), breast, and ovarian cancers [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] [bib_ref] HCMV Activates the IL-6-JAK-STAT3 Axis in HepG2 Cells and Primary Human Hepatocytes, Lepiller [/bib_ref] [bib_ref] The Implication of IL-6 in the Invasiveness and Chemoresistance of Ovarian Cancer..., Szulc-Kielbik [/bib_ref]. In addition, the proinflammatory environment triggered by HCMV will favor a lytic cycle through several mechanisms. First, proinflammatory cytokines such as TNFa and IL-6 will activate the major immediate early promoter (MIEP) of HCMV, favor viral transcription, and reactivate the virus from latency. Second, a proinflammatory environment will shift the maturation of monocytes toward macrophages, the latter being much more permissive than the former for HCMV replication [bib_ref] Human Cytomegalovirus Stimulates Monocyte-to-Macrophage Differentiation via the Temporal Regulation of Caspase 3, Chan [/bib_ref] [bib_ref] Human Cytomegalovirus Mediates Unique Monocyte-to-Macrophage Differentiation through the PI3K/SHIP1/Akt Signaling Network, Cojohari [/bib_ref]. In addition, HCMV-infected macrophages can produce anti-inflammatory cytokines such as TGF-beta and IL-10 that will deactivate the T-cell present in the vicinity of infected M1 macrophages, but also might shift macrophages toward an M2 phenotype. Although HCMV susceptibility was higher in M2 macrophages, HCMV establishes a productive and persistent infection in both types of macrophages [bib_ref] Human Cytomegalovirus Infection of M1 and M2 Macrophages Triggers Inflammation and Autologous..., Bayer [/bib_ref]. Thus, an exquisite balance between M1/M2 macrophage phenotype and HCMV replication kineticity could take place, especially with some HCMV strains, which are highly macrophage-tropic, such as the HCMV-DB strain [bib_ref] Bcl-3-Regulated Transcription from Major Immediate-Early Promoter of Human Cytomegalovirus in Monocyte-Derived Macrophages, Khan [/bib_ref].
HCMV might take advantage to buffer the proinflammatory microenvironment, diminish the antiviral immunity, and favor both viral spread and carcinogenesis. The production of G-protein-coupled receptors (GCRs) decoyed by HCMV, such as homologs US27, US28, UL33, and UL78, will hijack chemotactic factors, thus limiting the deleterious accumulation of the inflammatory cells at the viral infection site [bib_ref] Identification of the Human Cytomegalovirus G Protein-Coupled Receptor Homologue Encoded by UL33..., Margulies [/bib_ref]. Thus, HCMV US28, a homolog of the CCR1 gene, binds CC chemokines RANTES, monocyte chemoattractant protein-1 (MCP-1), MCP3, macrophage inflammatory protein-1 alpha (MIP-1α), and MIP-1β, in addition to the membrane-associated CX3C chemokine, fractalkine [bib_ref] HCMV-Encoded Chemokine Receptor US28 Mediates Proliferative Signaling through the IL-6-STAT3 Axis, Slinger [/bib_ref]. Although US28 activates phospholipase C and NF-κB, its role in cellular transformation and viral dissemination makes it a dual viral protein present during both productive and latent viral infection [bib_ref] Human Cytomegalovirus Chemokine Receptor Gene US28 Is Transcribed in Latently Infected THP-1..., Beisser [/bib_ref]. In such a context of proinflammatory immune control by HCMV, the v-IL10 immunosuppressive cytokine limits the mitogen-stimulated peripheral blood mononuclear cells' (PBMCs) proliferation and blocks proinflammatory cytokine synthesis in PBMCs and monocytes [bib_ref] Modulation of the Host Environment by Human Cytomegalovirus with Viral Interleukin 10..., Young [/bib_ref]. Of note, in such an immunocompromised microenvironment, HCMV enhances immune tolerance by inducing the transcription and release of TGF-β. TGF-β is well known to block antiviral IFN-γ and TNF-α cytokine production and cytotoxic effector activities of HCMV-specific Th1 cells and to be a key player in EMT, a critical event in the spread of transformed cells [bib_ref] Immune Landscape of CMV Infection in Cancer Patients: From "Canonical" Diseases toward..., El Baba [/bib_ref] [bib_ref] Human Cytomegalovirus Infection Induces Transcription and Secretion of Transforming Growth Factor Beta..., Michelson [/bib_ref]. The majority of the HCMV-encoded proteins and microRNAs (miRNAs) involved in preventing immune recognition are active mostly during the lytic phase [bib_ref] Human Cytomegalovirus Chemokine Receptor Gene US28 Is Transcribed in Latently Infected THP-1..., Beisser [/bib_ref].
HCMV shapes a microenvironment in the vicinity of infected cells where immunosuppressive cytokines such as vIL-10 and TGFb could also pave the way for oncogenic transformation. In the presence of vIL-10, HCMV infection of human cancer stem cells triggers macrophage reprogramming to an M2 phenotype in the tumor microenvironment, leading to the appearance of macrophages close to tumor-associated macrophages (TAMs), a hallmark of poor prognosis in many cancers [bib_ref] Tumor-Associated Macrophages: An Accomplice in Solid Tumor Progression, Chen [/bib_ref]. TAMs, the most abundant innate immune cells in tumors that exhibit an anti-inflammatory phenotype, are key players in cancer progression, metastasis, and resistance to therapy. A high TAM infiltration is generally associated with poor prognosis, but macrophages are highly plastic cells that can adopt either proinflammatory/antitumor or anti-inflammatory/protumor features in response to tumor microenvironment stimuli. Interestingly, upon CMV infection, macrophages undergo a morphological, immunophenotypic, and metabolic transformation process with features of stemness, altered migration, enhanced invasiveness, and provision of the cell cycle machinery for viral proliferation. Thus, CMV profoundly perturbs macrophage identity beyond established limits of plasticity and rewires specific differentiation processes, allowing viral spread and impairing innate tissue immunity. Interestingly, we recently identified a highly macrophage-tropic HCMV strain (DB strain), that shifted the infected macrophages toward an M2/TAM phenotype and upon acute infection transformed primary human mammary epithelial cells, further linking some HCMV clinical strains with the M2/TAM phenotype and oncogenesis [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref] [bib_ref] Bcl-3-Regulated Transcription from Major Immediate-Early Promoter of Human Cytomegalovirus in Monocyte-Derived Macrophages, Khan [/bib_ref].
Angiogenesis is a key player in cancer expansion, tumor relapse, and metastasis. HCMV activates the STAT3 pathway and the vascular endothelial growth factor (VEGF), which will further favor angiogenesis [bib_ref] The Role of Cytomegalovirus in Angiogenesis, Caposio [/bib_ref] [bib_ref] Human Cytomegalovirus IE1 Protein Disrupts Interleukin-6 Signaling by Sequestering STAT3 in the..., Reitsma [/bib_ref]. vIL-10 and US28 will also enhance cancer cell invasion and metastasis [bib_ref] CmvIL-10 Stimulates the Invasive Potential of MDA-MB-231 Breast Cancer Cells, Valle Oseguera [/bib_ref]. The enhanced survival of neutrophils and mononuclear cells in the tumor microenvironment will further promote oncogenesis via the activation of an angiogenic switch [bib_ref] Infiltrating Neutrophils Mediate the Initial Angiogenic Switch in a Mouse Model of..., Nozawa [/bib_ref] [bib_ref] Enhanced Neutrophil Activity Is Associated with Shorter Time to Tumor Progression in..., Rahbar [/bib_ref]. Altogether, HCMV will promote a pro-oncogenic environment enabling cancer progression through the activation of cancer stem cells, angiogenesis, invasion, and an EMT phenotype [bib_ref] Human Cytomegalovirus Infection Enhances Cell Proliferation, Migration and Upregulation of EMT Markers..., Teo [/bib_ref] [bib_ref] Human Cytomegalovirus Induces TGF-B1 Activation in Renal Tubular Epithelial Cells after Epithelial-to-Mesenchymal..., Shimamura [/bib_ref].
## Hcmv oncomodulation
The host immune system recognizes antigens resulting from alterations due to cancerous genetic and epigenetic instability [bib_ref] Genetic and Epigenetic Alterations as Hallmarks of the Intricate Road to Cancer, Macaluso [/bib_ref] [bib_ref] Cancer Genetics and Epigenetics: Two Sides of the Same Coin, You [/bib_ref] , which can resist immune clearance by prompting tolerance in the presence of tumor-associated inflammatory cells [bib_ref] Roles of the Immune System in Cancer: From Tumor Initiation to Metastatic..., Gonzalez [/bib_ref]. Thus, the tumor microenvironment takes place in an immune landscape that could be impacted by HCMV. The first paradigm that emerges to explain the role of HCMV in cancer is named oncomodulation, where the virus enhances the oncogenesis process already occurring in the transformed cells. Thus, the oncomodulation paradigm relies on the indirect role of CMV in cancer . In fact, the prevalence of HCMV is remarkably high in several cancer forms [bib_ref] Cytomegalovirus Is a Tumor-Associated Virus: Armed and Dangerous, Cobbs [/bib_ref] , but it is very difficult to know if the presence of HCMV is incidental due to the viral infection on the top of an already present tumor in the context of an immunosuppressive TME, or if the HCMV by itself can initiate and promote the cancer program, and thereby could be defined as a genuine oncovirus.
HCMV proteins and DNA are detected in over 90% of breast, colon, and prostate cancer, rhabdomyosarcoma, hepatocellular cancer, salivary gland tumors, neuroblastoma, and brain tumors [bib_ref] Cytomegalovirus Is a Tumor-Associated Virus: Armed and Dangerous, Cobbs [/bib_ref]. In addition, HCMV DNA was confirmed in up to 100% of breast cancer and 91% of sentinel lymph nodes samples from the metastatic group [bib_ref] High Prevalence of Human Cytomegalovirus Proteins and Nucleic Acids in Primary Breast..., Taher [/bib_ref]. A potential link between HCMV and metastatic cancer has been suggested since HCMV is detected in 98% of breast cancer-derived metastatic brain tumors [bib_ref] High Prevalence of Human Cytomegalovirus in Brain Metastases of Patients with Primary..., Taher [/bib_ref]. In metastatic breast cancer, the detection of HCMV results in shorter overall survival with the detection of HCMV DNA and transcripts in 92% and 80% of the used specimens, respectively [bib_ref] High Prevalence of Human Cytomegalovirus in Brain Metastases of Patients with Primary..., Taher [/bib_ref]. A negative correlation has been reported between the detection of IE1 and hormone receptor expression, indicating a potential link between HCMV and hormone receptor-negative breast cancer tumors [bib_ref] Söderberg-Naucler, C. Low Expression of Estrogen Receptor-α and Progesterone Receptor in Human..., Rahbar [/bib_ref]. The development of HCMV-related proliferative diseases may partially be ascribed to the ability of US28 to activate cyclooxygenase-2 (COX-2) [bib_ref] HCMV-Encoded Chemokine Receptor US28 Mediates Proliferative Signaling through the IL-6-STAT3 Axis, Slinger [/bib_ref] [bib_ref] The Human Cytomegalovirus-Encoded Chemokine Receptor US28 Promotes Angiogenesis and Tumor Formation via..., Maussang [/bib_ref]. COX-2 is frequently expressed in many types of cancers, exerting a pleiotropic and multifaceted role in the genesis or promotion of carcinogenesis and cancer cell resistance to chemo-and radiotherapy [bib_ref] Cyclooxygenase-2 in Cancer: A Review, Hashemi Goradel [/bib_ref]. Colon cancers, especially adenocarcinoma, often harbor high levels of HCMV, with IE1 and pp65 present in 82% and 78% of colorectal cancer samples with increased expression of Bcl-2 and COX-2 proteins, thus promoting colon cancer progression [bib_ref] Specific Localisation of Human Cytomegalovirus Nucleic Acids and Proteins in Human Colorectal..., Harkins [/bib_ref]. Hence, the use of COX-2-specific inhibitors, a subclass of nonsteroidal anti-inflammatory drugs (NSAIDs), and nonspecific COX-2 inhibitors such as aspirin reduce the cancer risk, potentiate antiangiogenic cancer therapy, and reduce metastasis in cancer patients [bib_ref] Cyclooxygenase-2 in Cancer: A Review, Hashemi Goradel [/bib_ref] [bib_ref] Reduction in Cancer Risk by Selective and Nonselective Cyclooxygenase-2 (COX-2) Inhibitors, Harris [/bib_ref]. The presence of HCMV was detected in preneoplastic and neoplastic prostate lesions, especially adenocarcinoma [bib_ref] High Prevalence of Human Cytomegalovirus in Prostatic Intraepithelial Neoplasia and Prostatic Carcinoma, Samanta [/bib_ref]. HCMV is also involved in disease progression in CNS tumors. Thus, HCMV is found in the majority of glioblastoma, and IE1 is expressed in glioma biopsy specimens in all grades [bib_ref] Human Cytomegalovirus Infection and Expression in Human Malignant Glioma, Cobbs [/bib_ref]. In addition, IE and late-HCMV proteins are detected in 100% and 92% of primary neuroblastoma samples, respectively, including neuroblastoma cells, which express high levels of stemness markers such as CD133 and CD44 [bib_ref] Frequent Detection of Human Cytomegalovirus in Neuroblastoma: A Novel Therapeutic Target?: Human..., Wolmer-Solberg [/bib_ref]. HCMV is also highly present in the primary medulloblastoma that expresses a high amount of IE and late-HCMV proteins, 92% and 73%, respectively [bib_ref] Detection of Human Cytomegalovirus in Medulloblastomas Reveals a Potential Therapeutic Target, Baryawno [/bib_ref]. Altogether, oncomodulation by HCMV has been observed in numerous adenocarcinoma and brain tumors. specific inhibitors, a subclass of nonsteroidal anti-inflammatory drugs (NSAIDs), and nonspecific COX-2 inhibitors such as aspirin reduce the cancer risk, potentiate antiangiogenic cancer therapy, and reduce metastasis in cancer patients [bib_ref] Cyclooxygenase-2 in Cancer: A Review, Hashemi Goradel [/bib_ref] [bib_ref] Reduction in Cancer Risk by Selective and Nonselective Cyclooxygenase-2 (COX-2) Inhibitors, Harris [/bib_ref]. The presence of HCMV was detected in preneoplastic and neoplastic prostate lesions, especially adenocarcinoma [bib_ref] High Prevalence of Human Cytomegalovirus in Prostatic Intraepithelial Neoplasia and Prostatic Carcinoma, Samanta [/bib_ref]. HCMV is also involved in disease progression in CNS tumors. Thus, HCMV is found in the majority of glioblastoma, and IE1 is expressed in glioma biopsy specimens in all grades [bib_ref] Human Cytomegalovirus Infection and Expression in Human Malignant Glioma, Cobbs [/bib_ref]. In addition, IE and late-HCMV proteins are detected in 100% and 92% of primary neuroblastoma samples, respectively, including neuroblastoma cells, which express high levels of stemness markers such as CD133 and CD44 [bib_ref] Frequent Detection of Human Cytomegalovirus in Neuroblastoma: A Novel Therapeutic Target?: Human..., Wolmer-Solberg [/bib_ref]. HCMV is also highly present in the primary medulloblastoma that expresses a high amount of IE and late-HCMV proteins, 92% and 73%, respectively [bib_ref] Detection of Human Cytomegalovirus in Medulloblastomas Reveals a Potential Therapeutic Target, Baryawno [/bib_ref]. Altogether, oncomodulation by HCMV has been observed in numerous adenocarcinoma and brain tumors. . HCMV oncomodulation and oncogenesis: two paradigms for one tumor. Oncomodulation will result from the infection of tumor cells by HCMV strains (named low-risk strains), which will enhance the clonogenic development of the tumor and result in a tumor with quite controlled cell homogeneity, limited metastasis, few relapses, sensitivity to chemotherapy and radiotherapy, and an overall of good prognosis. In contrast, direct oncogenesis will result from the infection of primary cells by HCMV strains (named high-risk strains), which could participate in giant cell cycling with the appearance of polyploid giant cancer cells (PGCCs), in a tumor with highly heterogeneous cancer cells, increased risk of metastasis, more relapses, chemoresistance, and, ultimately, poor prognosis. Of note, the two paradigms, oncomodulation versus oncogenesis, might not be mutually exclusive but could coexist in the same patient depending on the fitness of the HCMV strain, which might vary over time.
## Hcmv oncogenesis and the pgcc paradigm
Besides the oncomodulation paradigm, a direct oncogenic effect of HCMV has been recently forwarded [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref]. First, among the multiple cell types infected by HCMV, the stem cells are permissive to HCMV, and stem cells markers such as Thy-1 and plateletderived growth factor receptor alpha (PDGFRα) favor HCMV infection [bib_ref] THY-1 Cell Surface Antigen (CD90) Has an Important Role in the Initial..., Li [/bib_ref] [bib_ref] Platelet-Derived Growth Factor-α Receptor Activation Is Required for Human Cytomegalovirus Infection, Soroceanu [/bib_ref] [bib_ref] A Derivative of Platelet-Derived Growth Factor Receptor Alpha Binds to the Trimer..., Stegmann [/bib_ref] [bib_ref] Human Cytomegalovirus Glycoprotein Complex GH/GL/GO Uses PDGFR-α as a Key for Entry, Wu [/bib_ref]. Since stem cells, especially cancer stem cells (CSCs), play a critical role in tumor initiation, . HCMV oncomodulation and oncogenesis: two paradigms for one tumor. Oncomodulation will result from the infection of tumor cells by HCMV strains (named low-risk strains), which will enhance the clonogenic development of the tumor and result in a tumor with quite controlled cell homogeneity, limited metastasis, few relapses, sensitivity to chemotherapy and radiotherapy, and an overall of good prognosis. In contrast, direct oncogenesis will result from the infection of primary cells by HCMV strains (named high-risk strains), which could participate in giant cell cycling with the appearance of polyploid giant cancer cells (PGCCs), in a tumor with highly heterogeneous cancer cells, increased risk of metastasis, more relapses, chemoresistance, and, ultimately, poor prognosis. Of note, the two paradigms, oncomodulation versus oncogenesis, might not be mutually exclusive but could coexist in the same patient depending on the fitness of the HCMV strain, which might vary over time.
## Hcmv oncogenesis and the pgcc paradigm
Besides the oncomodulation paradigm, a direct oncogenic effect of HCMV has been recently forwarded [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref]. First, among the multiple cell types infected by HCMV, the stem cells are permissive to HCMV, and stem cells markers such as Thy-1 and plateletderived growth factor receptor alpha (PDGFRα) favor HCMV infection [bib_ref] THY-1 Cell Surface Antigen (CD90) Has an Important Role in the Initial..., Li [/bib_ref] [bib_ref] Platelet-Derived Growth Factor-α Receptor Activation Is Required for Human Cytomegalovirus Infection, Soroceanu [/bib_ref] [bib_ref] A Derivative of Platelet-Derived Growth Factor Receptor Alpha Binds to the Trimer..., Stegmann [/bib_ref] [bib_ref] Human Cytomegalovirus Glycoprotein Complex GH/GL/GO Uses PDGFR-α as a Key for Entry, Wu [/bib_ref]. Since stem cells, especially cancer stem cells (CSCs), play a critical role in tumor initiation, cellular transformation, tumor heterogeneity, and cancer dissemination through EMT and metastasis, it is worth questioning the direct role of HCMV in oncogenesis, potentially through the infection of stem cells and/or the appearance of cancer stem cells. In addition, stem cells can lose control over their self-growth and renewal, act as a cancer source, and become susceptible to oncogenesis in the presence of inflammation and altered DNA repair pathways; the latter has been observed frequently with some HCMV strains [bib_ref] Using or Abusing: Viruses and the Cellular DNA Damage Response, Lilley [/bib_ref]. HCMV could also favor stem cell survival, which would potentially sustain oncogenesis. Thus, IE1 protein promotes the presence of glioblastoma cancer stem cells through its induction of SRY-Box Transcription Factor 2 (SOX2), Nanog, Nestin, and octamer-binding transcription factor 4 (OCT3/4), key markers of stemness [bib_ref] Cytomegalovirus Immediate-Early Proteins Promote Stemness Properties in Glioblastoma, Soroceanu [/bib_ref] [bib_ref] Cytomegalovirus Infection Induces a Stem Cell Phenotype in Human Primary Glioblastoma Cells:..., Fornara [/bib_ref]. HCMV IE1 protein favors in glioblastoma cells the induction of transcription factors that are crucial for cancer stem cell persistence, cancer growth, and signaling pathways associated with the epithelial to mesenchymal (EMT) phenotype [bib_ref] Cytomegalovirus Immediate-Early Proteins Promote Stemness Properties in Glioblastoma, Soroceanu [/bib_ref] [bib_ref] Modulation of Oncogenic Phenotype in Human Glioma Cells by Cytomegalovirus IE1-Mediated Mitogenicity, Cobbs [/bib_ref]. Tumors are renowned as intricate systems that harbor heterogeneous cancer cells with distinctly diverse molecular signatures, sizes, and genomic contents. Among those various genomic clonal populations within the complex tumoral architecture are the polyploid giant cancer cells (PGCCs). Although described for over a century, PGCCs are increasingly being recognized for their prominent role in tumorigenesis, metastasis, therapy resistance, and tumor repopulation after therapy. A shared characteristic among all tumors triggered by oncoviruses is the presence of polyploidy [bib_ref] Polyploid Giant Cancer Cells, Herbein [/bib_ref]. Recently, our research team has highlighted the role of PGCCs as a critical factor following infection with some strains of HCMV, which can be described as oncogenic (or high-risk) HCMV strains (in addition to the oncomodulatory effect of the virus) [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref]. Two HCMV clinical strains isolated in our laboratory, named HCMV-DB and HCMV-BL, are capable of transforming primary human mammary epithelial cells (HMECs) into cytomegalovirus-transformed HMECs (CTH cells) and producing a "transcriptional profile" associated with DNA hypomethylation that results in the appearance of PGCCs in culture parallel to enhanced proliferation, activation of cancer stem cells, and EMT process [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] [bib_ref] The Transcriptome of Human Mammary Epithelial Cells Infected with the HCMV-DB, Moussawi [/bib_ref]. In addition, IE1 expression was detected in CMVtransformed HMECs (CTH) cells, including PGCCs, which also express embryonic stem cell markers [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref]. Xenografted NSG mice injected with CTH cells developed tumors that harbor HCMV DNA such as lncRNA4.9 gene, which displays a triple-negative basal-like phenotype. Thus, following acute infection of mammary epithelial cells (HMECs) with some HCMV strains (DB, BL), our team has shown that transformed cells appear in culture including PGCCs and are tumorigenic in xenografted NSG mice, indicating the direct involvement of HCMV in oncogenesis. The molecular mechanisms involved in HCMVinduced oncogenesis are multifactorial, including, among others, cellular stress, polyploidy, and genomic instability parallel to stemness appearance [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref] [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref]. HCMV gene products, including IE1 viral protein, could affect the pathways of p53 and Rb tumor suppressors and other pathways that are responsible for DNA repair parallel to increased expression of c-Myc in the transformed cells [bib_ref] Human Cytomegalovirus IE1 and IE2 Proteins Are Mutagenic and Mediate "Hit-and-Run, Shen [/bib_ref] [bib_ref] Viral Induction of Site-Specific Chromosome Damage, Fortunato [/bib_ref] [bib_ref] Human Cytomegalovirus UL76 Induces Chromosome Aberrations, Siew [/bib_ref]. Presuming the role of HCMV gene products in causing DNA damage directly and indirectly, and stimulating proliferation in stem cells, HCMV (or at least some high-risk HCMV strains) has the potential to initiate and promote tumor formation, especially following acute infection of epithelial cells, which will result in the appearance of adenocarcinoma of poor prognosis such as triple-negative breast cancer.
Nonetheless, our data indicate that the oncogenic potency of HCMV clinical strains varies between low-and high-risk strains [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] [bib_ref] Distinct Oncogenic Transcriptomes in Human Mammary Epithelial Cells Infected with Cytomegalovirus, Ahmad [/bib_ref] [bib_ref] The Transcriptome of Human Mammary Epithelial Cells Infected with the HCMV-DB, Moussawi [/bib_ref]. Only the high-risk HCMV strains can trigger the appearance of PGCCs [bib_ref] Immune Landscape of CMV Infection in Cancer Patients: From "Canonical" Diseases toward..., El Baba [/bib_ref] [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref]. HCMV-DB and HCMV-BL have been classified as high-risk strains since they showed a sustained transformation of acutely infected human mammary epithelial cells (HMECs) in vitro. These high-risk strains are characterized by elevated Myc expression, PI3K/Akt pathway activation, and p53 and Rb gene repression [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] [bib_ref] Human Cytomegalovirus Induces an Atypical Activation of Akt To Stimulate the Survival..., Cojohari [/bib_ref] [bib_ref] HCMV Modulation of Cellular PI3K/AKT/MTOR Signaling: New Opportunities for Therapeutic Intervention?, Altman [/bib_ref]. With regard to immune responses, Myc suppresses immune surveillance by modulating the expression of the innate immune regulator (CD47, also known as integrin-associated protein), the Treg activity, and the adaptive immune checkpoint, namely programmed death-ligand 1 (PD-L1) [bib_ref] Master Regulator of Immune Privilege, Casey [/bib_ref]. Further, PI3K/Akt hyperactivation observed with high-risk HCMV strains [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] induces immunosuppression and favors tumor initiation and progression via the activated Notch pathway [bib_ref] PI3K/Akt Cooperates with Oncogenic Notch by Inducing Nitric Oxide-Dependent Inflammation, Villegas [/bib_ref]. Loss of Rb leads to an increase in IL-6 production recently involved in the appearance of PGCCs in ovarian cancer [bib_ref] Retinoblastoma Inactivation Induces a Protumoral Microenvironment via Enhanced CCL2 Secretion, Li [/bib_ref]. Thus, high-risk HCMV strains are potentially involved in the oncogenesis process as described previously [bib_ref] The Human Cytomegalovirus, from Oncomodulation to Oncogenesis, Herbein [/bib_ref] [bib_ref] Polyploid Giant Cancer Cells, Stemness and Epithelial-Mesenchymal Plasticity Elicited by Human Cytomegalovirus, Nehme [/bib_ref] and might favor immune suppression in the tumor microenvironment further participating in tumor development. In addition, since the high-risk HCMV strains trigger the appearance of PGCCs, these latter are linked to undifferentiated heterogeneous tumors with poor prognosis; the immune environment of PGCCs needs definitively to be better characterized in the future.
HCMV infection fulfills all the hallmarks of cancer, including sustained proliferative signaling, deregulating cellular energetics, resisting cell death, favoring genome instability and mutation, inducing angiogenesis, activating invasion and metastasis, tumor-promoting inflammation, enabling replicative immortality, and avoiding immune destruction [fig_ref] Figure 3: HCMV fulfills all the hallmarks of cancer and thereby could modulate the... [/fig_ref] Viruses 2022, 14, 812 9 of 15 (reviewed in [bib_ref] Effects of Valganciclovir as an Add-on Therapy in Patients with Cytomegalovirus-Positive Glioblastoma:..., Stragliotto [/bib_ref]. Although HCMV fulfills all the hallmarks of cancer, high-risk HCMV strains could specifically activate invasion and metastasis, evade growth suppressors and contact inhibition, and favor the appearance of PGCCs in tumors of poor prognosis [fig_ref] Figure 3: HCMV fulfills all the hallmarks of cancer and thereby could modulate the... [/fig_ref]. Thus, the presence of PGCCs, one of the features of oncoviruses [bib_ref] Detection of Human Cytomegalovirus Proteins in Paraffin-Embedded Breast Cancer Tissue Specimens-A Novel, Touma [/bib_ref] , might be added as a new hallmark of cancer.
tumor development. In addition, since the high-risk HCMV strains trigger the appearance of PGCCs, these latter are linked to undifferentiated heterogeneous tumors with poor prognosis; the immune environment of PGCCs needs definitively to be better characterized in the future.
HCMV infection fulfills all the hallmarks of cancer, including sustained proliferative signaling, deregulating cellular energetics, resisting cell death, favoring genome instability and mutation, inducing angiogenesis, activating invasion and metastasis, tumor-promoting inflammation, enabling replicative immortality, and avoiding immune destruction [fig_ref] Figure 3: HCMV fulfills all the hallmarks of cancer and thereby could modulate the... [/fig_ref] (reviewed in [bib_ref] Effects of Valganciclovir as an Add-on Therapy in Patients with Cytomegalovirus-Positive Glioblastoma:..., Stragliotto [/bib_ref]. Although HCMV fulfills all the hallmarks of cancer, high-risk HCMV strains could specifically activate invasion and metastasis, evade growth suppressors and contact inhibition, and favor the appearance of PGCCs in tumors of poor prognosis [fig_ref] Figure 3: HCMV fulfills all the hallmarks of cancer and thereby could modulate the... [/fig_ref]. Thus, the presence of PGCCs, one of the features of oncoviruses [bib_ref] Detection of Human Cytomegalovirus Proteins in Paraffin-Embedded Breast Cancer Tissue Specimens-A Novel, Touma [/bib_ref] , might be added as a new hallmark of cancer.
## Antiviral therapy and immunotherapy used to fight hcmv-linked tumors
To confirm the role of HCMV in cancer development, glioblastoma (GBM) patients treated with valganciclovir, in addition to standard therapy, have demonstrated increased overall survival [bib_ref] Effects of Valganciclovir as an Add-on Therapy in Patients with Cytomegalovirus-Positive Glioblastoma:..., Stragliotto [/bib_ref]. Patients with newly diagnosed glioblastoma, receiving valganciclovir, had a longer median overall survival, a higher 2-year survival rate, and a longer median time-to-tumor progression compared to controls [bib_ref] Valganciclovir as Add-on to Standard Therapy in Glioblastoma Patients, Stragliotto [/bib_ref]. Thus, valganciclovir prolonged the median overall survival of patients with newly diagnosed glioblastoma and was safe to use [bib_ref] Valganciclovir as Add-on to Standard Therapy in Glioblastoma Patients, Stragliotto [/bib_ref]. Although antiviral therapy has not been found to definitively improve outcomes for patients with GBM, several studies have leveraged
## Antiviral therapy and immunotherapy used to fight hcmv-linked tumors
To confirm the role of HCMV in cancer development, glioblastoma (GBM) patients treated with valganciclovir, in addition to standard therapy, have demonstrated increased overall survival [bib_ref] Effects of Valganciclovir as an Add-on Therapy in Patients with Cytomegalovirus-Positive Glioblastoma:..., Stragliotto [/bib_ref]. Patients with newly diagnosed glioblastoma, receiving valganciclovir, had a longer median overall survival, a higher 2-year survival rate, and a longer median time-to-tumor progression compared to controls [bib_ref] Valganciclovir as Add-on to Standard Therapy in Glioblastoma Patients, Stragliotto [/bib_ref]. Thus, valganciclovir prolonged the median overall survival of patients with newly diagnosed glioblastoma and was safe to use [bib_ref] Valganciclovir as Add-on to Standard Therapy in Glioblastoma Patients, Stragliotto [/bib_ref]. Although antiviral therapy has not been found to definitively improve outcomes for patients with GBM, several studies have leveraged CMV by targeting CMV antigens using ex vivo expanded T cells or dendritic cell vaccines with some promising results [bib_ref] Three Times a Finding: Reproducibility of Dendritic Cell Vaccine Trials Targeting Cytomegalovirus..., Batich [/bib_ref]. Thus, immunotherapy used to fight HCMV infection in cancer is under development. To boost the immune system of patients, adoptive T-cell therapy using CMV-specific T cells, or dendritic cells pulsed with CMV pp65 RNA to vaccinate GBM patients, has been assessed. Thus, GBM tumors were infiltrated by CMV-specific T cells, and GBM killing was enhanced by dendritic cells pulsed with a CMV-derived peptide from the viral pp65 via pp65-specific T cells, parallel to improved survival of the patients. Interestingly, unlike healthy donors, large proportions of CMV-specific T cells in GBM patients were dysfunctional [bib_ref] Glioblastoma-mediated immune dysfunction limits CMV-specific T cells and therapeutic responses: Results from..., Weathers [/bib_ref]. Thus, immunotherapy based on vaccination and ex vivo T-cell expansion enhances the immune response against CMV-linked tumors and deserves further assessment (reviewed in [bib_ref] Cytomegalovirus and Immunotherapy: Opportunistic Pathogen, Novel Target for Cancer and a Promising..., Quinn [/bib_ref] [bib_ref] Insight for Immunotherapy of HCMV Infection, Long [/bib_ref].
# Conclusions
HCMV has been considered for decades as a herpesvirus involved mostly in asymptomatic or mild disease in immunocompetent individuals, which could lead to an important disease burden during congenital infection and in immunocompromised patients. Recently, the long-lasting effects of HCMV started to arise with its role in atherosclerosis, immunosenescence, and cancer. The oncomodulation paradigm could explain the accelerated progression of cancers when HCMV superinfection of the tumor occurs. Finally, very recently, a direct oncogenic effect of some HCMV strains, named high-risk strains, has been observed with cellular stress, appearance of PGCCs, stemness, and EMT, which could explain the appearance of aggressive cancers, especially adenocarcinoma, with poor prognosis, metastasis, and resistance to treatment. The oncomodulation and direct oncogenesis triggered by HCMV are not mutually exclusive and benefit from the same immunosuppressive tumor microenvironment. Overall, a better understanding of the complex role of HCMV in cancer and its immune landscape will lead to new therapeutic approaches that will curtail cancers, especially with poor prognosis.
Funding: This work was supported by grants from the University of Franche-Comté and the Région Franche-Comté (to G.H.).
## Conflicts of interest:
The author declares no conflict of interest.
[fig] Figure 1: HCMV rewires the immune landscape to favor both HCMV fitness and tumor growth through several mechanisms: increased tumor cell survival, tumor growth fueled by inflammation parallel to the appearance of an M2/Th2 immune landscape, and escape from immune surveillance by decreased specific CD4 + /CD8 + T-cell and NK responses. [/fig]
[fig] Figure 3: HCMV fulfills all the hallmarks of cancer and thereby could modulate the initiation and development of tumors. HCMV stimulates (green arrow) or blocks (red block) the molecular mechanisms involved in the hallmarks of cancer. The formation of PGCCs could be added as a new hallmark of cancer. High-risk HCMV strains might impact specifically some of the hallmarks of cancer where specified. [/fig]
|
Tetraparesis following an Anterior Circulation Stroke: A Case Report
The azygos anterior cerebral artery (AACA) is a large single anterior cerebral artery that supplies both medial territories of the anterior cerebral hemispheres. Occlusion of the AACA can result, therefore, in bifrontal infarction. We report a patient who suffered from a tetraparesis following a bilateral anterior cerebral artery territory infarction due to an occluded AACA and provide a brief review of the literature.
# Introduction
The azygos anterior cerebral artery (AACA), an anomaly of the circle of Willis, is a large single anterior cerebral artery (ACA) that supplies both medial territories of the anterior cerebral hemispheres. It has an incidence rate of 0.3-2.0% in adults [bib_ref] Thrombosis of the Azygos Anterior Cerebral Artery, De Sousa [/bib_ref] [bib_ref] Azygos anterior cerebral artery mimicking an anterior communicating artery aneurysm, Lightfoote [/bib_ref]. The AACA results from two vessels that instead of forming two ACA, fuse to one midline artery [bib_ref] Azygos anterior cerebral artery aneurysm associated with fenestration of the anterior cerebral..., Kobayashi [/bib_ref]. The AACA can be associated with other anomalies such as saccular aneurysm or arteriovenous malformation [bib_ref] Saccular aneurysm of the azygos anterior cerebral artery: three case reports, Huh [/bib_ref]. As the AACA is responsible for the perfusion of both anterior cerebral hemispheres, occlusion of the AACA results in bifrontal infarction [bib_ref] Thrombosis of the Azygos Anterior Cerebral Artery, De Sousa [/bib_ref] [bib_ref] Acute onset quadriplegia, Rajasekharan [/bib_ref]. We describe a patient who suffered from a tetraparesis following a bilateral ACA territory infarction due to an occluded AACA.
## Case description
A 62-year-old female patient complained of headache and slipped shortly after into a somnolent state with her eyes turned towards the right. Family anamnesis revealed that she had multiple falls during the last days.
On admission to the emergency unit, the patient showed eye deviation to the right and reduced vigilance. The initial working diagnosis consisted of a complex-partial status epilepticus, confirmed by electroencephalogram. Levitiracetam 750 mg and midazolam 4 mg intravenously were administered; subsequently, lacosamide and clonazepam were added for ongoing electroencephalographic epileptic activity.
A head MRI was performed due to prolonged leg-pronounced tetraparesis; imaging showed bilateral subacute ischemic lesions in the perfusion territory of the ACA. Time-offlight angiography revealed an unpaired ACA, in the form of an AACA, which was partially thrombosed. An arterio-arterial embolic event was considered as cause of the stroke [fig_ref] Figure 1: Axial MRI angiography showing an azygos anterior cerebral artery [/fig_ref]. Digital subtraction angiography showed two smaller aneurysms at the bifurcation of the AACA, which were left untreated after a careful risk-to-benefit evaluation. Aspirin and atorvastatin were initiated. The cardiac workup was unremarkable. The patient's clinical state remained unchanged, showing a psychomotor retardation, dysarthria, and tetraparesis (only a slight improvement of motor function of the right hand was noted). She was referred in a stabilized state to our clinic for rehabilitation and developed a progredient spastic paraparesis, while the initial upper extremity paresis recovered almost fully as well as the initial dysarthria. Cognitive performance improved drastically with a slight persistent cognitive slowing.
# Discussion
The literature reports few cases of AACA causing a stroke or an intracranial hemorrhage due to an associated ruptured aneurysm.
## Aaca and stroke
De Sousa et al. [bib_ref] Thrombosis of the Azygos Anterior Cerebral Artery, De Sousa [/bib_ref] reported a 63-year-old male patient with sudden lowering of his level of consciousness. The patient had a history of hypertension and diabetes. CT scan revealed an ischemic cerebral stroke in the typical irrigation territory of the bilateral ACA. Cerebral angiography evidenced an AACA with endoluminal thrombus, about 1 cm of its common origin, causing an ischemic stroke in the frontal lobe bilaterally and in the anterior regions of the corpus callosum. Rajasekharan and Deepak [bib_ref] Acute onset quadriplegia, Rajasekharan [/bib_ref] described a 50-year-old male patient with sudden onset of quadriplegia who had suffered a bilateral anterior cerebral artery infarct; imaging showed an AACA infarct as cause of the quadriplegia.
## Aaca and aneurysm
Huh et al. [bib_ref] Saccular aneurysm of the azygos anterior cerebral artery: three case reports, Huh [/bib_ref] presented three cases of AACA aneurysms among overall 781 cases of cerebral aneurysm. All three patients were elderly women presenting with subarachnoid hemorrhages. Small saccular aneurysms at the distal ends of the AACA were evidenced with cerebral angiography and CT-A. Lightfoote et al. [bib_ref] Azygos anterior cerebral artery mimicking an anterior communicating artery aneurysm, Lightfoote [/bib_ref] described the case of a 42-year-old man with headache of several months. CT scan showed a 4-mm densely enhancing soft tissue mass anterior to the suprasellar cistern with a punctate calcification in the expected position of the anterior communicating artery. A saccular unruptured aneurysm of the anterior communicating artery was suspected. However, intra-arterial digital subtraction angiography demonstrated a large (4 mm diameter, 1 cm long) AACA, divided into two pericallosal arteries and two callosomarginal arteries in their typical positions. No intracranial aneurysm was seen. A 56-year-old female with sudden onset of severe headache and vomiting was reported by Kobayashi et al. [bib_ref] Azygos anterior cerebral artery aneurysm associated with fenestration of the anterior cerebral..., Kobayashi [/bib_ref]. CT scan showed a severe subarachnoid hemorrhage in the longitudinal fissure of the cerebrum. Cerebral angiography revealed an AACA and a saccular aneurysm at the peripheral bifurcation of this artery.
# Conclusion
Quadriplegia following stroke is frequently due to posterior circulation occlusion (basilar strokes); however, anterior circulation occlusion due to AACA is an important differential diagnosis that equally deserves consideration.
# Statement of ethics
This material has not been published in whole or in part elsewhere; the manuscript is not currently being considered for publication in another journal; all authors have been personally and actively involved in substantive work leading to the manuscript, and will hold themselves responsible for its content.
# Disclosure statement
The authors declare that they have no conflict of interest.
[fig] Figure 1: Axial MRI angiography showing an azygos anterior cerebral artery (left). [/fig]
|
Clinical and neuropsychological changes after the disappearance of seizures in a case of transient epileptic amnesia
We encountered a female patient with late-onset temporal lobe epilepsy who presented with transient amnesia as the sole ictal manifestation, an accelerated rate of forgetting daily life events, and a retrograde memory deficit. We describe the memory function of the patient both before and after the administration of antiseizure medication. After the patient's seizures were controlled with antiseizure drugs, her neuropsychological memory performance scores showed improvement. We presumed that the disappearance of seizures was associated with a decrease in the accelerated rate of forgetting medication. However, her lost memories were not recovered after the seizures were controlled by antiseizure medication.
# Introduction
Memory disturbance is common among people with temporal lobe epilepsy (TLE). Transient epileptic amnesia (TEA) [bib_ref] Transient epileptic amnesia-a clinical update and a reformulation, Kapur [/bib_ref] is a subtype of TLE and is characterized by brief, recurrent episodes of transient amnesia during which other cognitive functions are preserved [bib_ref] Transient epileptic amnesia: a description of the clinical and neuropsychological features in..., Zeman [/bib_ref] [bib_ref] The syndrome of transient epileptic amnesia, Butler [/bib_ref]. The condition typically arises in middle and old ages. Most patients with TEA experience interictal memory difficulties such as an accelerated forgetting rate and isolated autobiographical memory deficit [bib_ref] Transient epileptic amnesia: a description of the clinical and neuropsychological features in..., Zeman [/bib_ref] [bib_ref] The syndrome of transient epileptic amnesia, Butler [/bib_ref]. The former is characterized by the normal acquisition and retention of memories over short periods of up to 30 min but abnormally fast forgetting over periods of days or weeks after the event. The phenomenon has been termed long-term amnesia or accelerated long-term forgetting [bib_ref] The syndrome of transient epileptic amnesia, Butler [/bib_ref] [bib_ref] Very long-term amnesia in association with temporal lobe epilepsy: evidence for multiple-stage..., Kapur [/bib_ref]. The latter is characterized by a patchy loss of memories of salient personal events, such as family events or holidays or weddings, from the remote past extending back over many years [bib_ref] The syndrome of transient epileptic amnesia, Butler [/bib_ref]. The performance of patients with these three types of memory deficit on standard neuropsychological memory tests that assess the retention of new memory after delays of 30-40 min is usually normal [bib_ref] Transient epileptic amnesia: a description of the clinical and neuropsychological features in..., Zeman [/bib_ref].
We encountered a woman with late-onset TLE who presented with transient amnesia as the sole ictal manifestation, an accelerated rate of forgetting of daily life events, and a retrograde memory deficit that specifically affected her autobiographical memory. We herein investigated her memory function both before and after the administration of antiseizure medication and discussed the findings about the relationships between her seizures during the study period and her memory function.
## Case presentation
Ms. A was a 67-year-old right-handed woman who was referred to us due to transient amnesia and memory disturbance. She was the product of a normal pregnancy and delivery, and her development was normal. Her family history was unremarkable. She had no history of head injury, neurological illness, or drug abuse. After graduating high school, she was employed as an office worker; after getting married, she became a homemaker. She was living with her husband, daughter and two granddaughters. Six months prior to her first attendance at our clinic, her family observed several episodes of transient amnesia of approximately 15 min in duration. These episodes were characterized by a sudden onset of disorientation regarding her location or her purpose for being at the location. Although impaired consciousness was absent and responsiveness was maintained, she had no recall for events during these attacks. She was unaware of her seizures. One morning, soon after getting up in her house, she asked her daughter where the switch for the heater was. She did not know the location of the switch for the electric floor heater. She looked at her husband and could not recognize him. On another occasion, when she and her family went on a day trip by, she suddenly repeated, "Why am I here?" She could respond when spoken to, although her responses were slightly superficial. At the same time as these episodes occurred, she also began to experience a baseline memory disturbance. She described accelerated forgetting as follows. She started to worry about her inability to remember what she had done approximately one month previously. When she consulted our clinic for a third time two months later, she did not remember that she had undergone a psychological examination at our clinic one month previously. She also described a patchy loss of remote autographical memories. Patchy memory loss of family travel and ceremonial occasions had occurred over the past 3 years. Even if she saw commemorative photos of the visits, she was not able to remember having traveled with her family one or two years previously. Her Mini-Mental State Examination and Hamilton Depression Rating Scale scores were 30 and 7, respectively. The results of the Wechsler Abbreviated Scale of intelligence (WAIS-III)and the Wechsler Memory Scale-Revised (WMS-R), are shown in [fig_ref] Table 1: Neuropsychological results [/fig_ref]. Her visual memory, general memory and delayed recall scores were in the lower ranges. Overall, her WMS-R score levels were lower in comparison to her WAIS-III score levels.
Interictal electroencephalography (EEG) revealed low-voltage spikes in the right temporal region on awake (Figs. 1 and 2) and during light sleep. Brain magnetic resonance imaging showed slight cerebral atrophy with no remarkable hippocampal atrophy. Interictal singlephoton emission computed tomography revealed an area of slight hypoperfusion in the right frontal and parietal regions.
Treatment with carbamazepine (50 mg, twice daily with a serum concentration of 3.1-4.5 μg/ml) was started focal epilepsy after the completion of the above-mentioned tests. Her seizures disappeared one month after starting carbamazepine. An EEG recording at one year after disappearance of seizures showed no epileptiform. One year after the disappearance of seizures, she no longer worried about forgetting daily life personal events. On repeat WAIS-IIIand WMS-R testingone year after disappearance of seizures, although there was no difference in her WAIS-III scores before and after the administration of the antiseizure medication, her visual memory, general memory and delayed recall scores in the WMS-R at one year after the disappearance of seizures improved overall in comparison to those before treatment [fig_ref] Table 1: Neuropsychological results [/fig_ref]. However, the lost memories about recent and remote personal events, such as family travel experiences, were not recovered.
# Discussion
Our patient showed recurrent episodes of isolated TEA as the sole manifestation of seizures and did not have any other features of seizures in TLE such as brief loss of awareness, olfactory or gustatory hallucinations and autonomic symptoms. She also had the characteristics of amnesia on waking and repeated questioning.
Some reports have suggested that TEA is sometimes the sole ictal manifestation of TLE [bib_ref] Transient epileptic amnesia-a clinical update and a reformulation, Kapur [/bib_ref] [bib_ref] Transient epileptic amnesia: a description of the clinical and neuropsychological features in..., Zeman [/bib_ref] [bib_ref] Epileptic amnesic attacks: benefit from antiepileptic drugs, Pritchard [/bib_ref] [bib_ref] Epileptic pseudodementia, Tatum [/bib_ref]. Ictal amnesia episodes are sometimes mistaken for episodes of transient global amnesia, transient ischemic attacks, or psychogenic amnesia [bib_ref] The syndrome of transient epileptic amnesia, Butler [/bib_ref]. The depression ratings of our patient were normal. She had no history of psychiatric disease and did not have any psychological problems.
Zeman et al. advocated the following diagnostic criteria for TEA [bib_ref] Transient epileptic amnesia: a description of the clinical and neuropsychological features in..., Zeman [/bib_ref] : 1) a history of recurrent witnessed episodes of transient amnesia; 2) cognitive functions other than memory were judged as being intact during typical episodes by a reliable witness; 3) sufficient evidence for a diagnosis of epilepsy (based on one or more of the following: a) epileptiform abnormalities on EEG, b) the concurrent onset of other clinical features of epilepsy, c) a clear-cut response to anticonvulsant medication). Butler et al. described patients with "pure" attacks of TEA, in which amnesia was the only ictal symptom [bib_ref] The syndrome of transient epileptic amnesia, Butler [/bib_ref]. They excluded cases where witness accounts were unavailable, unreliable, or indicated more extensive cognitive impairment during all attacks.
Although the presence of ictal EEG discharges during a typical event confirms the diagnosis of ictal amnesia [bib_ref] Epileptic pseudodementia, Tatum [/bib_ref] , we could not capture a typical event because of the infrequent occurrence of seizures in the present case. Our patient showed interictal low voltage spikes on her awake EEG. These spikes were different from benign epileptiform transients of sleep because of clear phase reversal in right temporal region and appearance on awake. Witness accounts of her family were available and reliable and she did not indicate more extensive cognitive impairment during all her attacks. She did not show the concurrent onset of other clinical features of epilepsy. From these facts, we concluded that her symptoms were broadly consistent with the diagnostic criteria for pure attacks of TEA.
Interictal epileptiform abnormalities on EEG in TEA were seen in about 40% and were localized over the temporal or fronto-temporal region [bib_ref] Recent insights into the impairment of memory in epilepsy: transient epileptic amnesia,..., Butler [/bib_ref]. Interictal EEG in our patient was similar to those reports. TEA is responsive to relatively low doses of antiseizure medication [bib_ref] The syndrome of transient epileptic amnesia, Butler [/bib_ref]. After commencing extremely low dose of sodium valproate monotherapy (daily dose = 100 mg), episodes ceased in patients with TEA [bib_ref] Autobiographical amnesia and accelerated forgetting in transient epileptic amnesia, Manes [/bib_ref]. Like these reports, treatment with low dose of antiseizure drug abolished the attacks in our patient.
When treating patients with epilepsy-associated memory problems, we need to consider the influence of antiseizure medication, because there is a possibility that the medication itself may be a confounding factor that affects the patient's memory function [bib_ref] The effects of various antiepileptic drugs on E-shockinduced amnesia in mice: dissociability..., Mondadori [/bib_ref] [bib_ref] Memory function in patients with epilepsy, Thompson [/bib_ref]. In the present study, the influence of antiseizure medication could be excluded because she was treatment-naïve.
Our patient not only had brief recurrent episodes of amnesia but also reported interictal accelerated forgetting and autobiographical amnesia. Her accelerated forgetting and the patchy impairment of episodic autographical memory were similar to the memory impairment that often occurs in TEA.
No previous reports have directly compared the pretreatment and posttreatment memory performance of patients in whom TEA was the sole manifestation of seizures. The findings regarding the relationship between seizure frequencies during the study period and memory function are controversial. Hendriks et al. noted that a high seizure frequency is particularly disruptive to the first encoding stage of the memory process [bib_ref] Relationships between epilepsy-related factors and memory impairment, Hendriks [/bib_ref]. Mameniskiene et al. reported that frequent seizures during the study period were related to poor long-term recall and that uncontrolled seizures can be a significant factor in the accelerated decay of memory [bib_ref] The decay of memory between delayed and long-term recall in patients with..., Mameniskiene [/bib_ref]. In addition, O'Connor et al. found that forgetting increased in conjunction with more frequent seizures and that this trend was reversed by antiseizure medication [bib_ref] Accelerated forgetting in association with temporal lobe epilepsy and paraneoplastic encephalitis, O'connor [/bib_ref]. In contrast, isolated seizures do not generally cause patients to forget material that they have recently learned, because there was no correlation between memory performance and seizure frequency [bib_ref] The effect of seizures on memory for recently learned material, Bergin [/bib_ref] [bib_ref] Accelerated forgetting in patients with epilepsy: evidence for an impairment in memory..., Blake [/bib_ref]. We presumed that the disappearance of epileptic seizures and the temporal lobe epileptiform abnormalities in our patient was associated with subjective memory improvement and the improvement in her WMS-R scores.
In our patient, even after seizures were controlled with antiseizure medication, the memories that lost in TEA, long-term anterograde and retrograde amnesia were not recovered. It is suggested that an irreversible change in the brain occurred due to repeated clinical and subclinical activity. Structures in both the hippocampal complex and neocortex play an important role in the establishment and maintenance of longterm episodic memory representation [bib_ref] Retrograde amnesia and memory consolidation: a neurobiological perspective, Squire [/bib_ref] [bib_ref] Consolidation and the hippocampal complex revisited: in defense of the multiple-trace model, Moscovitch [/bib_ref]. It remains unclear whether epileptic activity interferes with the memory consolidation, storage and retrieval processes. The possibility that recurrent seizures are responsible for the impairment of long-term memory consolidation has been raised by several authors. Kapur suggested that repeated burst of clinical and subclinical epileptiform activity over months and years may disrupt neocortically-based neural networks that act as storage or retrieval sites for long-term memory [bib_ref] Focal retrograde amnesia and the attribution of causality: an exceptionally benign commentary, Kapur [/bib_ref]. Manes et al. also suggested that extensive retrograde memory deficit is the cumulative effect of recurrent TLE with epileptic activity demonstrated on EEG [bib_ref] Focal autobiographical amnesia in association with transient epileptic amnesia, Manes [/bib_ref]. Recently a possible role for sleep in memory consolidation has been drawing attention. Slow wave sleep and rapid eye movement sleep have complementary function to optimize memory consolidation [bib_ref] The memory function of sleep, Diekelmann [/bib_ref]. Subclinical epileptiform activity during sleep may disrupt the hippocampal complex or neocortex, which act as the storage or retrieval sites for . Interictal monopolar EEG recording shows low voltage spikes predominantly in the right temporal region. . Interictal bipolar EEG recording shows low voltage spikes with phase reversal in the right temporal region. remote memory, and eliminate remote episodic memory. Further study is required to understand the functional mechanism underlying the accelerated forgetting and remote episodic memory impairment.
[fig] ⁎: Corresponding author at: Musashino Kokubunji Clinic, Kokubunji Honcho Crystal Building 3F, 4-1-9 Honcho, Kokubunji, Tokyo 185-0012, Japan. E-mail address: [email protected] (M. Sekimoto). //dx.doi.org/10.1016/j.ebcr.2017.01.002 2213-3232/© 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). [/fig]
[table] Table 1: Neuropsychological results. [/table]
|
The E3 ubiquitin ligase RNF40 suppresses apoptosis in colorectal cancer cells
Background: Colorectal cancer (CRC) is the fourth leading cause of cancer-related deaths worldwide, and deciphering underlying molecular mechanism is essential. The loss of monoubiquitinated histone H2B (H2Bub1) was correlated with poor prognosis of CRC patients and, accordingly, H2Bub1 was suggested as a tumor-suppressive mark. Surprisingly, our previous work revealed that the H2B ubiquitin ligase RING finger protein 40 (RNF40) might exert tumor-promoting functions. Here, we investigated the effect of RNF40 loss on tumorigenic features of CRC cells and their survival in vitro.Methods:We evaluated the effects of RNF40 depletion in several human CRC cell lines in vitro. To evaluate cell cycle progression, cells were stained with propidium iodide and analyzed by flow cytometry. In addition, to assess apoptosis rates, caspase 3/7 activity was assessed in a Celigo® S-based measurement and, additionally, an Annexin V assay was performed. Genomic occupancy of H2Bub1, H3K79me3, and H3K27ac was determined by chromatin immunoprecipitation. Transcriptome-wide effects of RNF40 loss were evaluated based on mRNA-seq results, qRT-PCR, and Western blot. To rescue apoptosis-related effects, cells were treated with Z-VAD-FMK.Results: Human CRC cell lines displayed decreased cell numbers in vitro after RNF40 depletion. While the differences in confluence were not mediated by changes in cell cycle progression, we discovered highly increased apoptosis rates after RNF40 knockdown due to elevated caspase 3/7 activity. This effect can be explained by reduced mRNA levels of antiapoptotic and upregulation of pro-apoptotic BCL2 family members. Moreover, the direct occupancy of the RNF40mediated H2B monoubiquitination was observed in the transcribed region of anti-apoptotic genes. Caspase inhibition by Z-VAD-FMK treatment rescued apoptosis in RNF40-depleted cells. However, knockdown cells still displayed decreased tumorigenic features despite the absence of apoptosis. Conclusions: Our findings reveal that RNF40 is essential for maintaining tumorigenic features of CRC cells in vitro by controlling the expression of genes encoding central apoptotic regulators.
# Background
Colorectal cancer (CRC) is one of the major causes of cancer-related deaths and incidence rates among people younger than 50 years have risen. Generally, the formation of CRC is a multistep process initiated by a hyperproliferation of the intestinal epithelium, leading to the formation of pre-cancerous polyps and, finally, adenomas and adenocarcinomas. This transition is based on the accumulation of molecular alterations, which are associated with the activation of oncogenes and the inactivation of tumor suppressor genes. These (epi-)genetic changes allow the cells to adopt characteristics referred to as "hallmarks of cancer" such as self-sufficiency in growth signals, limitless proliferative potential, and evasion of apoptosis.
The regulation of apoptosis, programmed cell death, is a tightly controlled process under physiological and pathological conditions and is essential for the clearance of cells displaying irreparable damage. The BCL-2 protein family plays a key role in controlling cell death and is subdivided into the pro-apoptotic (e.g., BAD, BAK, BAX, BCL-xS, BIM, and HRK) and anti-apoptotic factors (e.g., BCL-2, BCL-xL, and MCL1). These proteins control cytochrome-c release from the mitochondria, which results in the activation of a cascade of cysteine-aspartic proteases (caspases). Caspases mediate the cleavage of a broad variety of proteins responsible for normal cellular functions as well as cellular structure. As a consequence, a distinct morphological appearance, membrane blebbing, can be observed and the cell undergoes cell death. Since the evasion of apoptosis is a characteristic hallmark of cancer cells, induced cell death could cause the clearance of malignant cells and, therefore, presents a mechanism of targeting cancer cells. As reviewed by Baig and colleagues, several therapeutic approaches and agents are currently being tested clinically, including BH3 mimetics, which target proteins of the BCL-2 family in order to trigger apoptosis.
Interestingly, knockdown of the RING finger protein RNF20 was shown to downregulate the anti-apoptotic factor BCL-2 at the mRNA level in vitro. Generally, RNF20 and RNF40 form an obligate heterodimer with RNF40 which is recruited by the adaptor protein WW domain-containing adapter protein with coiled-coil (WAC) protein to the elongating RNA polymerase II large subunit following phosphorylation of serine 2 of the C-terminal heptapeptide repeat sequence. By exerting its E3 ligase activity, the RNF20/RNF40 complex was shown to monoubiquitinate histone H2B at lysine 120 (H2Bub1). It was proposed that H2Bub1 is associated with active transcriptional elongation by promoting the recruitment of the facilitates chromatin transcription (FACT) complex, which enhances chromatin accessibility and eases the passage of RNA polymerase II through the chromatin across the gene body. Importantly, H2Bub1 was described as a tumor-suppressive mark since reduced levels were associated with advanced tumor grade and poor survival in colorectal cancer patients. By extension, it has been postulated that the E3 ligases mediating the monoubiquitination of H2B also have a tumor suppressive function. Intriguingly, we previously demonstrated that the transient loss of RNF40 and accompanying the loss of H2Bub1 resulted in reduced proliferative potential of several CRC cell lines in vitro.
In this study, we used multiple approaches to investigate the mechanisms underlying these effects and have identified a previously unknown role for RNF40 and H2Bub1 in maintaining the expression of several anti-apoptotic genes. Together, these findings suggest that RNF40-mediated H2B monoubiquitination has a highly context-dependent function and may exert pro-tumorigenic functions in certain cellular contexts and thereby serve as a potential anti-cancer target.
# Methods
## Cell culture
Human colorectal cancer cell lines were grown in growth medium (HCT116, HT-29: McCoy's; RKO, SW48, SW837: Dulbecco's Modified Eagle's Medium/ F12) supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 μg/ml streptomycin at 37°C and 5% CO 2 . siRNA (GE Dharmacon siGENOME; nontargeting siRNA 5 [D-001210-05-20], RNF40 siRNAs [D-006913-01, -02, -03, -04]) transfections were performed using Lipofectamine® RNAiMAX (Invitrogen) according to the manufacturer's instructions. Twenty-four hours after siRNA transfection, cells were treated with 80 μM Z-VAD-FMK (Adooq) dissolved in DMSO or DMSO alone as a negative control for 48 h.
CRISPR/Cas9-mediated deletion of RNF40 pSpCas9(BB)-2A-GFP (PX458) was a gift from Feng Zhang (Addgene plasmid #48138)and was used to transfect cells with Cas9 along with the targeting guide RNAs (gRNAs). Guide RNAs were designed and checked for efficiency (http://cistrome.org/SSC) and specificity (http:// crispr.mit.edu). Subsequently, they were cloned in the plasmid and transfected into cells using the Lonza Nucleofector 2b device according to manufacturer's instructions (Kit T, program O-017). After 48 h of transfection, cells were sorted by flow cytometry (Cell Sorting Core Facility, Department of Hematology and Medical Oncology, University Medical Center Göttingen) and cells with highest GFP positivity were transferred as single cells into 96-well plates and propagated. For initial validation, cells were transfected with Lipofectamine 2000® (Life Technologies) according to the manufacturer's instructions. gRNA sequences are listed in Additional file 3: .
Validation of CRISPR/Cas9-mediated RNF40 knockout by PCR DNA was extracted from cells grown on 6-well plates by adding 300 μl lysis buffer (0.2% SDS, 100 mM Tris-HCl pH 8.5, 5 mM EDTA, 200 mM NaCl) and 40 μg proteinase K with incubation at 56°C and shaking overnight. DNA was precipitated with isopropanol and washed with 70% ethanol twice and re-constituted in water. DNA (100 ng) was amplified by PCR with 0.4 U Phusion polymerase (Thermo Scientific), 1× high fidelity buffer, 0.2 mM dNTPs, and 1 μM forward and reverse primers. The samples were heated at 98°C for 3 min followed by 35 cycles of 98°C for 30 s, 60°C for 30 s, and 72°C for 60 s. Finally, extension was performed for 10 min at 72°C. Forward primer: 5′-AGAAGCTCAGAACACGACGC-3′, reverse primer: 5′-TGCGTATCACATCCTCAGGG-3′. A PCR product of 1168 base pairs was expected in RNF40 wild-type cells and 312 bp in RNF40 knockout cells.
## Validation of crispr/cas9-mediated rnf40 knockout by immunofluorescence
Cells were grown on glass cover slips in 24-well plates for 24 h, and then washed three times with PBS and fixed using 4% paraformaldehyde in PBS for 10 min. Subsequently, cells were permeabilized with 0.5% Triton X-100 for 10 min and blocked with 3% BSA for 30 min prior to overnight incubation with RNF40 antibody (Sigma Aldrich, R9029) at 4°C. Cells were incubated for 1 h in secondary antibody conjugated to Alexa® Flour 594 (Life Technologies). Nuclei were stained with DAPI and coverslips were mounted on glass slides and left to dry at room temperature for 1 h and then stored at 4°C. Pictures were taken with a Zeiss LSM 510 Meta confocal 258 microscope.
## Cell characterization assays
Characterization assays were performed as previously described. Briefly, cells were seeded 24 h post siRNA transfection. Proliferation was tested by seeding 2000-5000 cells onto 96-well assay plates (Corning Life Sciences) and measuring the confluence daily using a Celigo® S cell imaging cytometer (Nexcelom Bioscience LLC). Clonogenic growth was assessed by seeding 500 cells onto a 6-well plate and waiting until macroscopically visible colonies were formed. Colonies were visualized by fixing cells in 4% PFA in PBS for 20 min, washed with PBS, and stained with 1% crystal violet in 20% ethanol for 20 min. Plates were washed with water and scanned. To test the migration potential, 100,000 HCT116 cells were seeded per PET track-etched membrane cell culture insert (8 μm, BD Bioscience) which was equilibrated with serum-free growth medium for 30 min. After 48 h, the inside of the inserts was scraped with a humidified cotton swab to remove non-migrated cells. Afterwards, migrated cells were fixed in 100% methanol for 10 min, stained with 0.1% crystal violet in 20% ethanol for 20 min, and rinsed with water. The stained area (%) covered by cells was quantified using FIJI.
## Qrt-pcr and western blotting
RNA was isolated using TRIzol (Invitrogen) and reverse transcribed using Moloney Murine Leukemia Virus reverse transcriptase (New England Biolabs) and random primers. Gene expression analysis was performed by quantitative real-time PCR using SYBR Green I (Roche Diagnostics). All expression values were normalized to 18S rRNA levels and corresponding primers are listed in Additional file 3: . For Western blot analyses, cells were lysed in RIPA buffer (1% NP-40, 0.5% sodium deoxycholate and 0.1% SDS dissolved in PBS) containing protease, phosphatase, and deubiquitinase inhibitors (1 ng/μL Aprotinin/Leupeptin, 10 mM β-glycerophosphate, 1 mM N-ethylmaleimide, 1 mM Pefabloc) and sonicated for 15 min. Proteins were denatured in 6× Laemmli buffer at 95°C for 5 min, separated by SDS-PAGE, and blotted onto nitrocellulose membranes. For RNA synthesis as well as protein isolation, floating and adherent cells were collected. Primary antibodies used for Western blot analyses are listed in Additional file 3: .
# Mrna-seq data analysis
Gene expression studies as well as Gene Set Enrichment Analysis (GSEA;following RNF40 depletion were performed as previously described(ArrayExpress accession number: E-MTAB-7197).
## Chromatin immunoprecipitation (chip)
To assess gene occupancy by H2Bub1, H3K27ac, and H3K79me3, SW837 cells were fixed in 1% formaldehyde in PBS for 20 min and quenched with 125 mM glycine for 5 min. Cells were scraped and washed with Nuclear Preparation Buffer (5 mM EDTA, 150 mM NaCl, 0.5% (v/v) NP-40, 50 mM Tris/HCl (pH 7.5), 1% (v/v) Triton X-100). The nuclear pellet was lysed using Lysis Buffer (20 mM EDTA, 150 mM NaCl, 1% (v/v) NP-40, 0.5% SDS, 0.5% (w/v) sodium deoxycholate, 20 mM sodium fluoride, 50 mM Tris/HCl (pH 8.0) and protease inhibitors) and sonicated for 20 cycles (30 s on/off). Samples were precleared with 50% Sepharose 4B (GE Healthcare) and incubated with the respective antibodies overnight (H3K27ac (196-050, Diagenode), H3K79me3 (C15310068, Diagenode) and H2Bub1 (5546, Cell Signaling)). Sepharose beads with Protein A (GE Healthcare) were added and incubated for 2 h, and immunocomplexes were washed twice with Wash Buffer (20 mM EDTA, 500 mM LiCl, 1% (v/v) 20 mM NaF, NP-40, 1% (w/v) sodium deoxycholate, 100 mM Tris (pH 8.5)) and twice with TE buffer. Next, samples were incubated with RNase A (Qiagen) in 10 mM Tris (pH 8.0) at 37°C for 30 min. For de-crosslinking, samples were incubated in 20 mM EDTA, 2% SDS, 100 mM Tris/HCl (pH 8), and 20 μg Proteinase K at 65°C for 4 h. Samples were extracted with Roti® phenol/chloroform/isoamylalcohol (Roth) and DNA was precipitated with ethanol. The corresponding input samples were used as controls.
## Chip-seq and data analysis
ChIP libraries were prepared using the NEBNext® Ultra DNA library preparation Kit (NEB) according to the manufacturer's instructions and samples were sequenced (single-end 50 bp) on a HiSeq2000 (Illumina). Quality control of fastq files was performed via FastQC (version 0.72), and reads were mapped to the human reference genome (GRCh38.p10) using Bowtie2 (version 2.3.4.2) with very sensitive presets in end-to-end mode on Galaxy. Reads were normalized to reads per kilobase per million (RPKM) by the bamCoverage tool (version 3.2.0.0.0) ignoring duplicates and extending for 125 base pairs. Using bigwigCompare (version 3.2.0.0.0), the signal of the input controls was subtracted from the respective data sets. Occupancy profiles were viewed by the Integrative Genomics Viewer (IGV 2.5.0). All ChIP-seq data have been deposited at ArrayExpress (http://www. ebi.ac.uk/arrayexpress, accession number: E-MTAB-7786).
# Cell cycle analysis
To analyze the cell cycle profile, adherent cells were trypsinized, washed with PBS, and fixed with 70% ice-cold ethanol overnight. Cell pellets were resuspended in PBS containing 10-20 μg/ml RNase (Macherey-Nagel GmbH & Co. KG) and 20-40 μg/ml propidium iodide (Sigma Aldrich). Samples were incubated at room temperature in the dark for 30 min. Cell cycle profiles were obtained using the Guava flow cytometry system (Millipore) using the same gate settings for each cell line. Three biological replicates with each two technical replicates were processed in three independent experiments.
## Annexin v assay
Floating cells were collected and adherent cells were trypsinized, filtered, and washed with ice-cold PBS. After centrifugation, 100,000 cells were resuspended in 100 μl binding buffer (0.01 M HEPES pH 7.4, 0.14 M NaCl, 2.5 mM CaCl 2 ). Propidium iodide (Sigma Aldrich) and Annexin V-FITC (Southern Biotech) were added (1:100) and incubated at room temperature for 15 min. Four hundred microliters of binding buffer was added, and apoptosis analyses were performed using the Guava flow cytometry system (Millipore) using the same gate settings for each cell line. The experiment was performed with three biological replicates per condition, each with two technical replicates in three independent experiments.
## Viastain tm apoptosis assay
To evaluate caspase 3/7 activity, a Celigo® S-based ViaStain TM (Nexcelom Bioscience) apoptosis assay was carried out. An endpoint analysis was performed according to the manufacturer's instructions 72 h after the siRNAmediated knockdown of RNF40.
## Statistical analyses
All graphs in this study were made with GraphPad Prism version 5.04 (GraphPad Software, Inc.). Statistical analysis was performed using the one-way analysis of variants (ANOVA) and Tukey post hoc test (α = 0.05, *p ≤ 0.05, **p ≤ 0.01,***p ≤ 0.001).
# Results
## Rnf40 knockdown reduces growth and tumorigenic features of crc cell lines
Our previous studies suggested that RNF40 might be required for the pro-proliferative behavior of colorectal cancer cells in vitro. To test the general impact of RNF40 reduction on CRC cells, we examined the morphology of HCT116 and four additional CRC cell lines not previously tested following siRNA-mediated knockdown of RNF40 (siRNF40) compared to control cells (siControl). The knockdown efficiency was defined at the mRNA and protein levels (Additional file 1: . As we previously observed in other CRC cell lines, while the gross morphology was not affected by RNF40 knockdown 72 h post-transfection , Celigo® S-based quantification revealed that the confluence was significantly reduced in siRNF40 cells . We next evaluated the effect of RNF40 knockdown on the clonogenic capacity of the investigated cell lines. Indeed, the ability of single cells to form colonies was significantly reduced following RNF40 knockdown . In addition, the migration potential of HCT116 RNF40 knockdown cells was significantly reduced compared to the control cells . In summary, these findings demonstrate that the loss of RNF40 leads to reduced cell confluence as well as decreased tumorigenic features of all CRC cell lines tested.
## Permanent crispr/cas9-mediated rnf40 deletion is lethal in crc cells in vitro
In order to validate our observations upon depletion of RNF40 by siRNA, we sought to establish a permanent RNF40 deletion in CRC cells using CRISPR/Cas9-mediated gene editing. For this purpose, we used an approach to delete both exons 3 and 4 by using two guide RNAs (gRNAs) targeting the intron regions between exon 2 and 3 and exons 4 and 5, respectively. This approach was chosen since it specifically mirrors the conditional Rnf40 knockout mouse model previously described by targeting the homologous regions and exons of the human RNF40 gene. Importantly, the resulting loss of exons 3 and 4 will lead to a frameshift of any resulting spliced transcript, thereby ensuring the lack of functionality of any resulting protein (Additional file 2:. To validate the effectiveness of RNF40 deletion, we transfected SW480 cells with RNF40 gRNAs cloned in a GFP-fused Cas9 plasmid (pSpCas9(BB)-2A-GFP; PX458, Addgene) using Lipofectamine 2000® and examined RNF40 expression by immunofluorescent staining. Interestingly, 24 h after transfection, we observed that GFP-expressing cells were devoid of RNF40 expression, validating the efficiency of our approach (Additional file 2:. We subsequently utilized electroporation to obtain a higher efficiency of transfection, and after 48 h, we sorted single, highly positive GFP-expressing cells by flow cytometry into single wells in a 96-well plate format to obtain single cell clones. While we were able to readily sort cells based on high GFP expression (Additional file 2:, not a single resulting clone (out of 42) was positive for a homozygous deletion of RNF40 (Additional file 1: . Alternative approaches utilizing gRNAs directly targeting different exons (exons 2, 3, 6, and 9) also failed to produce any homozygous clones (data not shown). This pattern of lethality was also observed in MDA-MB-231 breast cancer cells, which also failed to produce any knockout clones even after screening ca. 150 clones (Additional file 2:. Together, these data implied that the complete and permanent loss of RNF40 is lethal in CRC cells in vitro. Therefore, in order to further investigate the molecular mechanisms underlying this lethality, all subsequent experiments were performed using a transient, siRNA-mediated knockdown of RNF40.
## Rnf40 loss does not affect cell cycle progression in crc cells
To elucidate the mechanisms underlying the decreased cell number following RNF40 depletion (or deletion), we evaluated gene expression signatures enriched following RNF40 depletion by performing Gene Set Enrichment Analysis (GSEA) of our previously published mRNA-seq dataset in HCT116 cells. Previous analyses uncovered a regulation of cell cycle-associated gene signatures related to cell division, cell cycle phase transition, and DNA replication as being negatively enriched in HCT116 RNF40 knockdown cells. Consequently, we sought to evaluate the effects of RNF40 knockdown and, consequently, the reduction on H2Bub1 levels (Additional file 1: on the cell cycle. For this purpose, we tested the proportions of RNF40 loss reduces the tumorigenic properties of CRC cells in vitro. Functional assays were performed using HCT116, HT-29, RKO, SW48, and SW837 cells after siRNA-mediated RNF40 knockdown in three independent experiments. a The morphology of CRC cell lines was not affected 72 h after RNF40 knockdown. Scale bar, 1 mm. b Cellular confluence was measured 72 h post-transfection using a Celigo® S device and decreased cell numbers were detected following RNF40 depletion (n = 6). Mean ± SEM, Students t test. c The clonogenic capacity of CRC cells was determined by quantifying well areas covered by crystal violet-stained colonies formed from single cells (n = 3). Upon loss of RNF40, the cells formed fewer colonies. Mean ± SEM, Students t test. d Representative images of the colony formation assay. e Migration potential was significantly decreased in HCT116 cells with reduced RNF40 levels. Mean ± SEM, Students t test adherent HCT116 and SW837 control and RNF40 knockdown cells in the respective phases of the cell cycle using propidium iodide staining. Surprisingly, we did not detect any significant changes in the flow cytometry profiles as well as in proportions of cells in the respective cell cycle phases after RNF40 knockdown in either cell line. Thus, despite the regulation of cell cycle-associated genes, the distribution of cells in the respective cell cycle phases was not appreciably affected by RNF40 depletion.
## Rnf40 knockdown cells display apoptotic features
Since only negligible effects of RNF40 knockdown on the cell cycle could be observed, we aimed to elucidate further processes underlying the decreased confluence of siRNF40 cells. We observed high numbers of nonadherent cells in the medium upon RNF40 loss . Thus, after seeding the same number of cells for control and knockdown cells, we quantified the respective cell populations. Indeed, the relative number of floating cells was four to sixfold increased 72 h after RNF40 depletion . Moreover, in our HCT116 GSEA data, we observed an enrichment of apoptosis-associated genes after the loss of RNF40 .
## Rnf40 depletion promotes apoptosis
Next, we aimed to investigate in more detail whether reduced RNF40 levels promote apoptosis in CRC cells in vitro. SW837 and HCT116 siControl and siRNF40 cells were stained with Annexin V-FITC and propidium iodide 72 h after siRNA transfection to measure the proportion of late apoptotic cells. Indeed, we found an increase in the fraction of apoptotic cells following RNF40 depletion in both cell lines . To verify these findings, we performed a Celigo® S-based ViaStain TM (Nexcelom Bioscience) apoptosis assay. Here, cells were incubated with a specific probe (coupled to a fluorescent dye) which is capable of binding to DNA. Upon caspase 3/7-mediated apoptosis, the complex is cleaved, thereby releasing the dye which translocates to the nucleus where it generates a fluorescent signal. In support of our Annexin V staining, we were able to detect an increase in apoptotic cells following RNF40 knockdown using this approach . While only a slight but significant increase in caspase 3/7 activity was detected in SW837 cells, the increase of apoptotic cells in HCT116 cells was nearly fivefold. Together, our data clearly demonstrate that reduced expression of RNF40 promotes apoptosis of CRC cells.
## Rnf40 knockdown increases apoptosis by affecting bcl-2 family members
To gain mechanistic insights into how the siRNAmediated depletion of RNF40 promotes apoptosis, we analyzed our HCT116 mRNA-seq data for the expression of several pro-and anti-apoptotic genes. Interestingly, several members of the BCL-2 family were differentially regulated at the mRNA level in RNF40-depleted HCT116 cells. While anti-apoptotic factors such as BCL2, The proportions of cells in the respective cell cycle phases were determined in HCT116 and SW837 cells and no significant differences were detected between RNF40 control and knockdown conditions. Mean ± SEM MCL1, and BIRC5 were downregulated upon the loss of RNF40, pro-apoptotic genes (e.g., BAX, BCL2L11, and HRK) were upregulated. These findings could be confirmed by qRT-PCR which confirmed that BCL2L11 mRNA levels were increased up to threefold in RNF40 knockdown cells. As described earlier by our group, H2Bub1 occupancy is highly associated with gene expression and active histone marks such as H3K27ac and is required for downstream methylation of lysine 79 of histone H3. Consistent with the positive role of RNF40-mediated H2B monoubiquitination in controlling gene expression, ChIPseq analysis in SW837 cells revealed high occupancy of H2Bub1 and H3K79me3, as well as H3K27ac, on the RNF40-dependent anti-apoptotic genes MCL1 and BIRC5. Moreover, as detected by Western blot analysis, BIM (encoded by BCL2L11) was substantially increased in siRNF40 cells, a finding which was confirmed in SW837 cells. Moreover, we assessed the levels of poly (ADP-ribose) polymerase (PARP) and detected increased amounts of cleaved PARP. Generally, PARP is a substrate of active caspases and, therefore, high cleaved PARP levels confirmed our ViaStain TM results that caspase-mediated apoptotic processes are increased in RNF40-depleted cells. In summary, our findings suggest that RNF40 suppresses apoptosis by maintaining the expression of anti-apoptotic BCL-2 family members via H2B monoubiquitination on the transcribed region of these genes.
## Inhibition of caspase activity maintains viability of rnf40 knockdown cells
Based on the elevated caspase 3/7 activity and the increase in cleaved PARP upon RNF40 knockdown, we hypothesized that the apoptotic phenotype of these cells can be rescued by inhibiting caspases. For this purpose, 24 h after siRNA transfection, HCT116 and SW837 cells were treated with 80 μM Z-VAD-FMK (herein referred to as Z-VAD) for 48 h. As expected, in DMSO-treated control cells, the loss of RNF40 resulted in lower cell numbers and the presence of a high proportion of detached cells . In contrast, after Z-VAD treatment, less floating cells were visible. To quantitate the proportions of adherent and non-adherent cells, we counted all cells floating in the medium . While Z-VAD treatment showed only minimal effects on siControl cells, the number of viable cells was significantly elevated by Z-VAD in RNF40-depleted cells. These findings were confirmed by Western blot analysis where PARP cleavage could be substantially reduced, Cell morphology and differentially regulated genes suggest apoptotic processes in RNF40 knockdown cells. a A higher proportion of nonadherent cells was observed after the siRNA-mediated knockdown of RNF40. Scale bar, 1 mm. b The same number of cells was seeded prior to the knockdown, and 72 h later, the number of floating and attached cells was counted in three independent experiments. The relative number of non-adherent cells was significantly higher after the loss of RNF40. Mean ± SEM, Student's t test. c Gene set enrichment analysis indicated increased apoptotic processes after RNF40 knockdown in HCT116 cells while BIM levels remained unchanged . Together, these results support the findings that RNF40 loss enhances apoptosis in CRC cells due to increased caspase activity.
## Rnf40 loss reduces tumorigenic properties independent of apoptosis regulation
After determining that the apoptotic phenotype of RNF40-depleted cells could be rescued by caspase inhibition, we aimed to determine whether RNF40 may have independent roles in controlling apoptosis and other tumorigenic features. Thus, we performed additional functional assays in HCT116 cells in the presence of Z-VAD to inhibit caspase activity and subsequent cell death. First, we carried out colony formation assays to test the capacity of single cells to form colonies. As expected, siRNF40 cells formed fewer colonies than the controls; however, this phenotype was not rescued after Z-VAD treatment . Next, we evaluated the migration capacity, which was significantly reduced after the loss of RNF40 . After restoring cell growth using Z-VAD, we were indeed able to slightly increase the number of migrated cells following RNF40 knockdown which was, however, still significantly lower compared to wild-type controls. In summary, even after rescuing the increased apoptosis in RNF40 knockdown The loss of RNF40 promotes apoptosis in CRC cells. a HCT116 and b SW837 cells were stained with Annexin V-FITC and propidium iodide and analyzed using the Guava flow cytometry system (Millipore) 72 h after control or RNF40 knockdown. Three independent experiments were performed with three biological replicates per condition each with two technical replicates. The relative number of apoptotic cells was significantly increased in both cell lines after RNF40 knockdown. Mean ± SEM, Student's t test. c Gated cells of the flow cytometry-based Annexin V assay in HCT116 and d SW837 cells. Red frames indicate gated late apoptotic cells. e Caspase 3/7 activity was measured using a Celigo® S-based ViaStain TM (Nexcelom Bioscience) apoptosis assay and confirmed increased apoptosis after the loss of RNF40 in HCT116 and f SW837 cells. Mean ± SEM, Student's t test. g Representative images of the ViaStain TM assay indicating DAPI-stained nuclei (blue) and cells with caspase 3/7 activity (green) cells, RNF40 depletion still reduces additional tumorigenic properties of CRC cells in vitro.
# Discussion
A tight regulation of apoptotic signaling is central to maintaining tissue homeostasis and a loss of the balance between cell death and survival and evasion of apoptosis is a common feature of cancer cells. In this study, we were able to demonstrate that the knockdown of the E3 ligase RNF40 and, consequently, the loss of H2Bub1 results in reduced tumorigenic potential and increased apoptosis in human colorectal cancer cells in vitro. b Verification of mRNA-seq results in HCT116 cells using qRT-PCR in three independent experiments. RNF40-depleted HCT116 cells display decreased mRNA levels of the anti-apoptotic factors MCL1 and BIRC5 while pro-apoptotic genes BAX and BCL2L11 were upregulated (n = 3); c these findings were confirmed in SW837 cells. Mean ± SEM, Student's t test. d ChIP-seq profiles show the occupancy of H2Bub1, H3K27ac, and H3K79me3 on the anti-apoptotic genes MCL1 and BIRC5 in SW837 cells. Input signals were subtracted. e In three independent experiments, protein was harvested from HCT116 and SW837 cells and analyzed using Western blot. In both cell lines, RNF40 loss elevated the protein amounts of BIM (encoded by BCL2L11) and the apoptosis marker cleaved PARP H2Bub1 loss was also connected to cell death in previous studies. For instance, knockdown of a RNF20/ RNF40 homolog in Caenorhabditis elegans resulted in germ cell apoptosis. In addition, Duan and colleagues investigated the relevance of these E3 ligases in mitosis and spindle assembly. They discovered that a loss-of-function of RNF20/RNF40 resulted in G2/M arrest and apoptosis by downregulating the motor protein Eg5 in the human breast cancer cell line MCF-7. Moreover, in TNFα-treated mammary epithelial cells, the knockdown of RNF20 reduced the expression of the anti-apoptotic factor BCL-2 in vitro. All these results suggest an anti-apoptotic function of RNF40. While we did not observe any cell cycle changes in SW837 and HCT116 cells, we were able to confirm that loss of RNF40 induces apoptosis.
Using qRT-PCR and ChIP-seq, we demonstrated that the loss of RNF40, presumably via decreased H2Bub1 gene occupancy, modulates the expression levels of various BCL-2 family members, thereby promoting apoptosis. Interestingly, several reports described the interplay between BCL-2 members and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), and it was suggested that NF-κB activity enhances the expression of anti-apoptotic genes. This postulation is in agreement with our previous findings that the knockdown of RNF40 delayed the nuclear translocation of the NF-κB member RelA after TNFα treatment. Consequently, the expression of NF-κB target genes was decreased after the loss of either RNF40 or RNF20. Future studies will provide further insight into the interplay between RNF40, BCL-2 family members, and NF-κB.
Besides the role of RNF40 in apoptosis, we investigated the effects of RNF40 depletion on other tumorigenic features of CRC cells in vitro. Our earlier findings implicated that RNF40 exerts oncogenic functions and promotes aggressive growth of CRC cells in vitro. Here, we were able to verify in additional CRC cell lines that RNF40 knockdown reduces proliferation as well as clonogenic and migration potential. To rule out the possibility that increased apoptosis rates underlie the Caspase inhibition rescues apoptosis in RNF40 knockdown cells. Twenty-four hours after knockdown, HCT116 and SW837 cells were treated with 80 μM Z-VAD for 48 h to inhibit caspases in three independent experiments. a Morphologically, the appearance of siRNF40 cells was highly similar to the controls after Z-VAD treatment as exemplarily shown for HCT116 cells. Scale bar, 1 mm. b The percentage of adherent and detached cells was assessed. The proportion of dead cells was significantly reduced in HCT116 cells and a similar trend was observed in c SW837 RNF40 knockdown cells. Mean ± SEM, one-way ANOVA. c Western blot analysis revealed that Z-VAD treatment decreased the levels of cleaved PARP, while BIM remained unchanged after RNF40 knockdown in HCT116 and e SW837 cells reduced cell growth of RNF40-depleted cells, we rescued apoptosis-associated effects by inhibiting caspases using Z-VAD. Notably, even after rescuing RNF40depleted cells from apoptosis by Z-VAD treatment, cells displayed decreased clonogenic and migratory potential. Therefore, we were able to support the hypothesis that RNF40 exerts oncogenic functions in CRC. This finding is not in agreement with other correlative studies which described that the loss of H2Bub1 in colorectal tumors was linked to advanced tumor grades and poor survival. However, our own data demonstrated that colorectal tumors display a high degree of heterogeneity in H2Bub1 levels, while RNF40 amounts were constant in cancer lesions, thereby suggesting that RNF40 may play additional roles in CRC apart from the monoubiquitination of H2B. However, the underlying mechanisms of such an uncoupling remain to be elucidated but may be related to upstream signaling and/or downstream deubiquitination of H2Bub1.
# Conclusion
Together, our findings suggest that RNF40 exerts protumorigenic functions in colorectal cancer in vitro by increasing clonogenic potential as well as by suppressing apoptosis. This occurs by the regulation of BCL-2 family members, which can either suppress or promote apoptosis. Further experimental procedures will help to elucidate further underlying molecular mechanisms employed by RNF40 to reduce apoptotic signaling and to determine whether RNF40 may serve as a therapeutic target for inducing apoptosis in colorectal cancer cells.
## Additional files
Additional file 1: . Reduction of RNF40 and H2Bub1 in siRNF40 CRC cell lines. . The siRNA-mediated knockdown of RNF40 was verified at the mRNA level using qRT-PCR (A) and on protein level using western blot (B) in three independent experiments 72 h after RNF40 knockdown cells with restored growth capacity show decreased tumorigenic properties. The effect of RNF40 knockdown in combination with Z-VAD treatment was tested in three independent experiments. a The clonogenic potential of HCT116 siRNF40 cells was reduced compared to wild-type controls. This phenotype was not rescued after Z-VAD-mediated caspase inhibition (n = 2). Mean ± SEM, one-way ANOVA. b Migration potential was tested using a trans-well migration assay for 48 h. Z-VAD treatment increased the proportion of migrating cells under RNF40 knockdown conditions which was still significantly lower compared to wild-type cells (n = 3). Mean ± SEM, one-way ANOVA transfection. Mean ± SEM, Students t test. (C) The knockdown of RNF40 resulted in decreased H2Bub1 levels 72 h after siRNA transfection. (ZIP 7432 kb) Additional file 2:. RNF40 knockout is lethal in colorectal cancer cells in vitro. (A) Diagram showing the approach to establish a permanent RNF40 knockout SW480 cells. Two gRNAs were targeted at intron regions before exon 3 and after exon 4 leading to a frameshift and a non-functional protein product. (B) Immunofluorescence assay showing the lack of RNF40 detection (red) when expressing GFP (green) which indicates successful transfection with the Cas9 construct. Scale bar: 10 μM. |
Late combination shows that MEG adds to MRI in classifying MCI versus controls
a b s t r a c tEarly detection of Alzheimer's disease (AD) is essential for developing effective treatments. Neuroimaging techniques like Magnetic Resonance Imaging (MRI) have the potential to detect brain changes before symptoms emerge. Structural MRI can detect atrophy related to AD, but it is possible that functional changes are observed even earlier. We therefore examined the potential of Magnetoencephalography (MEG) to detect differences in functional brain activity in people with Mild Cognitive Impairment (MCI) -a state at risk of early AD. We introduce a framework for multimodal combination to ask whether MEG data from a resting-state provides complementary information beyond structural MRI data in the classification of MCI versus controls. More specifically, we used multi-kernel learning of support vector machines to classify 163 MCI cases versus 144 healthy elderly controls from the BioFIND dataset. When using the covariance of planar gradiometer data in the low Gamma range (30-48 Hz), we found that adding a MEG kernel improved classification accuracy above kernels that captured several potential confounds (e.g., age, education, time-of-day, head motion). However, accuracy using MEG alone (68%) was worse than MRI alone (71%). When simply concatenating (normalized) features from MEG and MRI into one kernel (Early combination), there was no advantage of combining MEG with MRI versus MRI alone. When combining kernels of modality-specific features (Intermediate combination), there was an improvement in multimodal classification to 74%. The biggest multimodal improvement however occurred when we combined kernels from the predictions of modality-specific classifiers (Late combination), which achieved 77% accuracy (a reliable improvement in terms of permutation testing). We also explored other MEG features, such as the variance versus covariance of magnetometer versus planar gradiometer data within each of 6 frequency bands (delta, theta, alpha, beta, low gamma, or high gamma), and found that they generally provided complementary information for classification above MRI. We conclude that MEG can improve on the MRI-based classification of MCI.
# Introduction
Alzheimer's disease (AD) is an age-related neurodegenerative disorder and a major challenge for healthcare and social care due to its high prevalence and costs. Early detection of AD is critical for treatment and prevention, and this requires a robust biomarker that can identify the disease in cases such as Mild Cognitive Impairment (MCI) [bib_ref] Early diagnosis of Alzheimer's disease: is MCI too late?, Petersen [/bib_ref]. Such biomarkers may also provide disease progression monitoring in clinical trials. Neuroimaging techniques offer a range of potential biomarkers of structural, metabolic and functional changes in the brain related to AD and MCI [bib_ref] Maintenance, reserve and compensation: the cognitive neuroscience of healthy ageing, Cabeza [/bib_ref] [bib_ref] Neuroimaging in dementia, Tartaglia [/bib_ref] [bib_ref] Building better biomarkers: brain models in translational neuroimaging, Woo [/bib_ref].
Magnetic Resonance Imaging (MRI) is such a key technique, which can be tuned to various tissue properties. The most common of these is [bib_ref] Exploring the brain network: a review on resting-state fMRI functional connectivity, Tsvetanov [/bib_ref]. Furthermore, the slow haemodynamic response means that fMRI is largely blind to neuronal dynamics above 0.1 Hz.
More direct measures of neural activity can be obtained by Electroencephalography (EEG) and Magnetoencephalography (MEG), which measure the electromagnetic fields produced by dendritic dipoles within active neurons [bib_ref] Use of magnetoencephalography (MEG) to study functional brain networks in neurodegenerative disorders, Stam [/bib_ref]. These can be sampled at a resolution of milliseconds, revealing a rich repertoire of neural dynamics, including oscillatory rhythms that occur at frequencies between 2 and 100 Hz, such as "alpha " (8-12 Hz) and "gamma " (30 + Hz), some of which have also been implicated in AD [bib_ref] Alzheimer's disease: the state of the art in resting-state magnetoencephalography, Engels [/bib_ref] [bib_ref] Deep-MEG: spatiotemporal CNN features and multiband ensemble classification for predicting the early..., Giovannetti [/bib_ref] [bib_ref] The role of magnetoencephalography in the early stages of alzheimer's disease, López-Sanz [/bib_ref] [bib_ref] A comprehensive review of magnetoencephalography (MEG) studies for brain functionality in healthy..., Mandal [/bib_ref] [bib_ref] Altered generation of spontaneous oscillations in Alzheimer's disease, Osipova [/bib_ref] [bib_ref] Magnetoencephalographic evaluation of resting-state functional connectivity in Alzheimer's disease, Stam [/bib_ref] [bib_ref] Stability of spectral estimates in resting-state magnetoencephalography: recommendations for minimal data duration..., Wiesman [/bib_ref]. MEG offers an advantage over EEG in that the magnetic fields are less distorted and smoothed by the brain-skull interface than are electric fields, resulting in higher spatial resolution . We, therefore, focus on MEG measures of neural activity (during rest) to see if they provide information for MCI classification that is complementary to the structural information in sMRI.
The general advantage of multimodal integration (combining information from more than one neuroimaging technique) has been appreciated for many years, on the assumption that each modality reveals information about somewhat different aspects of the underlying neural circuity [bib_ref] A parametric empirical Bayesian framework for the EEG/MEG inverse problem: generative models..., Henson [/bib_ref] [bib_ref] BOLD and EEG signal variability at rest differently relate to aging in..., Kumral [/bib_ref] [bib_ref] Functional connectivity of EEG is subject-specific, associated with phenotype, and different from..., Nentwich [/bib_ref] [bib_ref] Combining anatomical, diffusion, and resting state functional magnetic resonance imaging for individual..., Schouten [/bib_ref] and consistent with findings that combining multiple modalities can improve AD classification [bib_ref] EEG and MRI data fusion for early diagnosis of Alzheimer's disease, Patel [/bib_ref] [bib_ref] Multimodal EEG, MRI and PET data fusion for Alzheimer's disease diagnosis, Polikar [/bib_ref]. For example, using various different machine learning techniques, [bib_ref] EEG and MRI data fusion for early diagnosis of Alzheimer's disease, Patel [/bib_ref] and [bib_ref] Multimodal EEG, MRI and PET data fusion for Alzheimer's disease diagnosis, Polikar [/bib_ref] have demonstrated that combining EEG and sMRI improved diagnosis of AD versus controls compared to using individual modalities alone. However, the EEG data in these studies were collected during an auditory oddball paradigm, rather than the more common resting-state used here. [bib_ref] Multimodal EEG-MRI in the differential diagnosis of Alzheimer's disease and dementia with..., Colloby [/bib_ref] is the only study we could find that combined resting-state EEG and sMRI, but they focused on distinguishing relatively late cases of AD versus Lewy-body dementia, rather than distinguishing MCI versus healthy controls, as done here (a study by [bib_ref] A comparison of resting state EEG and structural MRI for classifying Alzheimer's..., Farina [/bib_ref].
Neuroimaging produces many measurable properties (or "features ") from each participant, such as ∼100,000 voxels in an sMRI image or ∼1000,000 timepoints in ∼300 sensors in MEG, which normally exceed the number of participants (typically ∼100). Identifying which features are important for classifying AD, therefore, benefits from machine learning techniques [bib_ref] From estimating activation locality to predicting disorder: a review of pattern recognition..., Wolfers [/bib_ref] , such as kernel-based approaches. A kernel is a square matrix containing a measure of the similarity between every pair of participants in their feature values. Classification based on kernels rather than raw data is robust and efficient for highdimensional pattern classification. Here we used Multiple Kernel Learning (MKL) [bib_ref] Multiple kernel learning algorithms, Gönen [/bib_ref] , which optimises the weighting of kernels from each modality, which is generally better than simply concatenating features across modalities, particularly when the modalities differ in the number of features and/or those features are incommensurate (such as volume in mm 3 for sMRI versus magnetic field power in fT 2 for MEG) [bib_ref] A multimodal multiple kernel learning approach to Alzheimer's disease detection, Donini [/bib_ref] [bib_ref] Biomagnetic biomarkers for dementia: a pilot multicentre study with a recommended methodological..., Hughes [/bib_ref] [bib_ref] Predicting progression from mild cognitive impairment to Alzheimer's dementia using clinical, MRI,..., Korolev [/bib_ref] [bib_ref] Improving alzheimer's disease classification by combining multiple measures, Liu [/bib_ref] [bib_ref] Structured sparsity regularized multiple kernel learning for Alzheimer's disease diagnosis, Peng [/bib_ref] [bib_ref] Identification of MCI individuals using structural and functional connectivity networks, Wee [/bib_ref] [bib_ref] Multi-kernel learning with Dartel improves combined MRI-PET classification of Alzheimer's disease in..., Youssofzadeh [/bib_ref] [bib_ref] Multimodal classification of Alzheimer's disease and mild cognitive impairment, Zhang [/bib_ref]. Adding kernels to the classification model is also a better way to accommodate potentially confounding variables (such as the age of MCI and Control cases) than is the more common approach of first adjusting the data (features) for those variables [bib_ref] Controlling for effects of confounding variables on machine learning predictions, Dinga [/bib_ref] [bib_ref] How to control for confounds in decoding analyses of neuroimaging data, Snoek [/bib_ref].
Most importantly, we compared results from combining modalities at three different stages: early, intermediate and late [fig_ref] Figure 1: Schematic showing the three different combination stages used here [/fig_ref]. By Early combination, we refer to the simple concatenation of the features of each modality, after normalizing them by their standard deviation across participants (i.e., to unit-less quantities with comparable numerical range). By Intermediate combination, we refer to the typical MKL approach of optimizing the weighting of kernels derived from the features of each modality (or confound). By Late combination, we refer to the application of MKL to kernels derived from the class predictions after classifiers are run on each modality separately. The latter is closer to the "ensemble learning " philosophyand the "stacking " approach used by.
# Materials and methods
## Participants
We included resting-state MEG and T1-weighted structural MRI data from the BioFIND dataset, which includes individuals with Mild Cognitive Impairment (MCI) and Healthy Elder Controls (HEC) from two sites: the MRC Cognition & Brain Sciences Unit (CBU) in Cambridge, England, and the center for Biomedical Technology (CTB) in Madrid, Spain. The CBU controls were recruited from the CamCAN sample ( www.cam-can.org ) who are screened to be healthy, i.e., have MMSE (and indeed ACE-R) scores above conventional cut-offs, as well as other screening described in CamCAN paper [bib_ref] The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study protocol: a cross-sectional,..., Shafto [/bib_ref]. The CTB controls had a full neuropsychological assessment to confirm normal cognition, and the same type of MRI assessment as that done in the MCI group, i.e., a radiologist reported MRIs as normal.
In general, the MCI diagnosis was determined with the intermediate probability according to the National Institute on Aging-Alzheimer Association criteria [bib_ref] The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from..., Albert [/bib_ref] , i.e., given by a clinician based on clinical and cognitive tests, self-and informant-report, and in the absence of full dementia or obvious other causes (e.g., psychiatric). More specifically, for the CBU data, individuals were diagnosed with MCI after referral for symptoms, mainly memory problems (i.e., they were not derived by screening of cognitively asymptomatic people). The diagnosis was made in a regional memory clinic, including ACE/ACER and MMSE tests as standard, with a significant deficit in memory domain tests. PET and fluidic Biomarkers were not used as standard, although all had structural brain imaging (usually MRI in the clinic unless contraindicated, when a CT is occasionally used) and clinical follow-up in support of the diagnosis. By definition of MCI, sufferers had functional independence at the time of diagnosis. sMRI was used to exclude other pathologies and to identify features consistent with MCI/AD pathology (e.g. MTL atrophy without mass lesion, high vascular burden).
For the CTB data, the diagnosis of MCI was based on criteria from [bib_ref] Revised NIA-AA criteria for the diagnosis of Alzheimer's disease: a step forward..., Frisoni [/bib_ref]. It was based on a mixture of clinical and quantitative approaches, but MMSE score was not the sole criterion. The diagnosis required an impairment in the memory domain and/or other cognitive functions, but a "memory complaint" was not required, so some participants showed awareness of their cognitive problems and others did not. Indeed, for CTB data participants were not aware of their current cognitive status. Therefore, they did not have to show cognitive complaints, but memory/cognitive failures were reported by close companions. The only biomarkers used systematically were brain morphology (from MRI) and APOE. As a requirement for the diagnosis criteria, MCI cases were able to still perform their daily living activities. MRI (T1, T2 and/or FLAIR) was used to rule out a vascular disorder, and any other type of neurological disease (i.e., tumor, stroke, infection) that could better explain the cognitive symptoms.
After excluding 15 cases without an MRI, and 2 with MRIs with dental artifacts, there were 163 HEC and 144 MCI datasets. A summary of sample characteristics is reported in [fig_ref] Table 1: Characteristics of BioFIND participants with clean MRIs [/fig_ref] , which includes variables that could affect MCI status (such as education) or could affect brain activity in general (such as time of day of testing) or could affect the MEG data specifically (such as head motion or distance from sensors). A small number of missing values were imputed using the mean from nonmissing values for each variable: 5 missing values for MMSE (1.5% of cases, all MCI), 12 missing values for Education (3.7% of cases; 11 MCI, 1 HEC) and 26 missing values for mean and standard deviation of motion (8.0% of cases; 21 MCI and 5 HEC). We used MATLAB's "knnimpute ", applied across the whole sample, which uses the value from the nearest neighbor based on Euclidean distance. The imputed data are available in the tab-separated value file "participants-imputed.tsv " on the GitHub repository ( https://github.com/delshadv/MRI _ MEG _ Combination ).
## Confounds
There were more HEC cases from the CBU site and more MCI cases from the CTB site, but the number of males/females was close to be matched across the two groups. A two-sample T-test confirmed that, on average, the MCI group had lower MMSE scores, as expected from their clinical diagnosis, T(305) = 9.63, p < 0.001. However, they also had fewer years in education, T(305) = 6.41, p < 0.001, and were slightly older on average, T(305) = 2.09, p < 0.05, which may confound any group differences in MRI and/or MEG. While none of the other variables in [fig_ref] Table 1: Characteristics of BioFIND participants with clean MRIs [/fig_ref] differed significantly between the two groups, T's < 1.46, p > 0.14, it is possible that combinations of them could predict MCI status above chance. We, therefore, included all of them as potential confounds (COFs), except MMSE. The reason we did not include MMSE as a confound is that this cognitive measure informs the MCI diagnosis, so would be circular (biased) to use as a predictor.
## . meg preprocessing
MEG data were acquired while participants were seated inside a magnetically shielded room (MSR). The CBU MSR is made by Imedco and uses single layer mu-metal plates, while the Madrid MSR is made by Vaccumschmelze and has two layers (for further technical details, see.
The raw data were de-noised using signal space separation (SSS) implemented in MaxFilter 2.2.12 (Elekta Neuromag) to suppress environmental noise [bib_ref] Presentation of electromagnetic multichannel data: the signal space separation method, Taulu [/bib_ref]. For more details of maxfiltered data and parameters applied to MEG data in BioFIND, please see. The max-filtered (and raw) data are available here: https://portal.dementiasplatform.uk/AnalyseData/AnalysisEnvironment .
The max-filtered data were read into using the SPM12 toolbox ( http://www.fil.ion.ucl.ac.uk/spm/ ;. The minimum duration of resting-state data across participants was 120 s, so the first two minutes of data were used for all participants. A recent study by [bib_ref] Stability of spectral estimates in resting-state magnetoencephalography: recommendations for minimal data duration..., Wiesman [/bib_ref] showed that spectral properties of intrinsic brain activity can be robustly estimated from such short segments of restingstate data. (Indeed, when we repeated the current analyses with 3 min of data, excluding the five participants who had less than 3 min of data, the results were virtually identical). The precise pre-processing steps are provided in the preproc_meg.m script in the GitHub repository (see above link). In brief, the 120 s of data was down-sampled to 500 Hz and band-passed filtered from 0.5 to 98 Hz (via a high-pass filter fol-lowed by a low-pass filter). The filter type is Butterworth IIR filter with the order of 5 using spm_eeg_filter.m function. The continuous data were then epoched into 2s segments, and bad epochs were marked using the OSL automatic artefact detection ( https://ohba-analysis.github.io/osldocs/ ). The number of bad epochs ( M = 4.66 for MCI and M = 4.03 for HEC, out of 60 total) did not differ significantly between groups, T(305) = 1.59, p = 0.11. Non-bad epochs were then concatenated again and bandpass filtered within each of six frequency bands.
Unfortunately, not all participants had EOG or ECG recorded, in order to provide objective measures of artifacts. However, note that eyes were closed in all cases, which will abolish blinks and minimize eye motion, while the dominant power in ECG is typically around 1 Hz, which is below the lowest frequency analyzed here. Moreover, the first author conducted a brief visual inspection of data from all participants. This revealed residual some spikes, particularly in the high-frequency bands (low gamma and high gamma). After band-pass filtering, these spikes were detected and corrected using MATLAB's filloutliers function (using the 'median' option with a 'ThresholdFactor' of 3 for detection, and the "clip " method for replacement). The number of such spikes (outliers) did not differ significantly between groups in any frequency band, T's < 1.33, p 's > 0.19, except Delta, T(305) = 2.60, p = 0.01, and was always fewer than 1% of samples.
## Mri preprocessing and feature extraction
The de-faced, T1-weighted scans were processed in SPM's DARTEL-VBM pipeline [bib_ref] Voxel-based morphometry-the methods, Ashburner [/bib_ref] , as implemented in the preproc_mri.m script in the GitHub repository. Each MRI was first segmented into gray matter (GM) and white matter (WM) probability maps. These GM and WM images were then warped to an average template for the sample using diffeomorphic warping in the DARTEL toolbox [bib_ref] A fast diffeomorphic image registration algorithm, Ashburner [/bib_ref] , and this template transformed to MNI space. These transformation parameters were then applied to each participant's GM image, modulated so as to preserve local GM volume, and the GM values resampled into MNI space together with a spatial smoothing by a 1 mm FWHM isotropic Gaussian kernel to remove interpolation artifacts. Finally, 110 MRI features were used for classification, representing the mean across voxels within the 110 anatomical ROIs of the Harvard-Oxford Atlas [bib_ref] Gyri of the human neocortex: an MRI-based analysis of volume and variance, Kennedy [/bib_ref] [bib_ref] MRI-Based topographic parcellation of human cerebral white matter and nuclei II. Rationale..., Makris [/bib_ref]. 1 In case the HOA atlas is too coarse to reveal anatomical changes in MCI, Supplementary Section S6 shows results using GM volume in all 390,189 voxels as features instead.
## Meg feature extraction
We focused on sensor level features, rather than reconstructing the sources of the MEG data. Firstly, it is unclear whether source reconstruction provides additional information for classification with real data. Since sensor data are a linear combination of source data (mixed through a "forward model ",, there no degrees of freedom is gained when estimating source amplitudes. Having said this, when classifying on the basis of non-linear functions of the data, such as the (co)variance (power) features used here, this equivalence between sensor and source features is lost [bib_ref] Manifold-regression to predict from MEG/EEG brain signals without source modeling, Sabbagh [/bib_ref]. It is true that an accurate forward model (which accommodates differences in head position and anatomy) helps align features across participants, and simulations show improvements when estimating sources , though in practice there are always errors in the forward model and noise in the data that prevent perfect alignment. Indeed, there are normally more sources than sensors, rendering the source estimation problem ill-posed, and requiring additional assumptions to regularize the solution. Secondly, and perhaps more importantly, since an sMRI is necessary to construct an accurate head model, for the present purposes of comparing MRI and MEG classification, we did not want information from the MRI to contaminate the MEG features. The MEG data come from one magnetometer (MAG) and two, orthogonal planar gradiometers (GRD) at each of 102 locations above the head (i.e., 306 channels in total). MAGs measure the component of the magnetic field that is perpendicular to the sensor, whereas GRDs estimate the spatial derivative of the magnetic fields in two orthogonal directions in the plane of the sensor (which is roughly parallel to the scalp). By taking the spatial derivative, GRDs are less sensitive to distant sources, such as environmental noise, and so have a higher signal-to-noise ratio for signals close by, i.e., in the superficial cortex. By contrast, MAGs are more sensitive to deeper signals in the brain, but also more susceptible to noise. The best way to combine GRD and MAG data is a matter of contention because they have different SI units [bib_ref] Choice of magnetometers and gradiometers after signal space separation, Garcés [/bib_ref] , but by using separate kernels for each, we do not need to combine them directly.
We focused on the second-order moments of MEG data, i.e., the data covariance across sensors (which is incidentally what most source localization methods are based on). The variances are related to the signal power in each sensor, whereas the covariances are related to the cross-spectral power. Note that sensor covariances capture aspects of both brain activity and connectivity (and can only be attributed solely to connectivity between sources after adjusting for field spread, i.e., linear mixing by the forward model,. This meant 102 features for MAG variance and 204 features for GRD variance, both corresponding to the spatial distribution of power across the scalp, with 5151 covariance features for MAGs and 20,706 covariance features for GRDs. Note however that our preprocessing of the MEG data (during the MaxFilter stage above) includes a dimension reduction to 69 and 66 components for MAG and GRD respectively, so the rank of the covariance matrices is less. This does not matter for our classification results, since when we reduced the dimensionality further, using principal component analysis (PCA) [bib_ref] Principal component analysis, Wold [/bib_ref] to calculate the number of components needed to explain 95% of the total variance in the MEG features across participants, the pattern of classification results was hardly changed (see Supplementary Section S4). Note that, to avoid bias, PCA was applied in the training set and the projection was then applied to the test set (see mkl_ens_nestedcv.m and mkl_ens_n.m in the paper's GitHub repository).
Each . As described in preprocessing section, the frequency transform is applied to continuous data (without epoching). Note that we did not relativize power in each frequency band to the total power (across all frequencies), e.g. by normalizing the time-series before estimating power [bib_ref] Biomagnetic biomarkers for dementia: a pilot multicentre study with a recommended methodological..., Hughes [/bib_ref]. While such normalization allows for differences in overall signal strength owing to the proximity of the head (brain) to the sensors, the danger of normalizing power in this way is that it could also remove true power differences between MCI and HEC. We, therefore, used absolute power, but by including the mean distance between the brain and sensors as a confound, were able to make some allowance for different head positions (after squaring, since the magnetic field strength falls off with at least the square of distance).
## Multimodal classification
We compared three stages of combining MEG and MRI data: 1. Early combination: features from MRI and MEG are normalized and then concatenated and fed to a single classifier; 2. Intermediate combination: features from MRI and MEG are projected to kernels and MKL then 2 There are data-driven alternatives, for example, to adjust the sensor covariance to more accurately reflect source covariance using Riemannian Embedding , which could be tried in future and compared with present results.
forms an optimal kernel from a function of the original kernels that maximises multimodal classification accuracy; 3. Late combination: MRI and MEG features are fed to one or more classifiers whose continuous-valued outputs (a prediction related to the probability of membership of each class or distance to decision boundary) are then combined in an optimal way, again using MKL [bib_ref] Multiple kernel learning algorithms, Gönen [/bib_ref]. According to a common taxonomy of classifier ensemble methods, our Late combination is a type of "stacked generalization ", since the outputs of the individual classifiers for each modality are treated as inputs to a meta classifier (here EasyMKL) [bib_ref] Stacked generalization, Wolpert [/bib_ref]. It is also worth mentioning that in pattern recognition, Early combination is categorized as "feature level " combination whereas Late combination is classified as "decision level " combination. Using Intermediate or Late combinations can be a natural way to control the confounds (COF) effect, namely by adding one or more kernels (see Discussion). A comparison schematic of these approaches is presented in [fig_ref] Figure 1: Schematic showing the three different combination stages used here [/fig_ref].
Kernel methods project the data features into matrices (kernels) that represent the similarity of the feature vectors between every pair of observations (here, participants), and optimize classification performance based on these kernels (thus each kernel in the present case was a 307 ×307 matrix, regardless of the number of features per modality). Kernels are the basis of several types of classifiers such as Support Vector Machines (SVM) [bib_ref] Support-vector networks, Cortes [/bib_ref] and MKL ( Gönen and Alpayd ı n, 2011 ). Let { x , y } L be the training set, where i = 1,2,…L is the number of training examples. Then, a linear MKL learns a coefficient vector that weights each kernel k to produce an optimal kernel K according to:
[formula] ( , ′ ) = ∑ =1 η ( , ′ ) with ∑ =1 η = 1 , η ≥ 0 [/formula]
where n = 1,2,…N is the number of base kernels. Note that we used linear kernels (as well as a linear combination), which is recommended when the number of features is larger than the number of participants, i.e., when there is insufficient data to fit more complex nonlinear kernels. The decision function of an MKL problem which is the distance to the decision boundary for classification models can be then expressed in the form:
[formula] ( ) = ∑ =1 * K ( , ) + * [/formula]
Where * and * are some coefficients to be learned from training examples. Learning these coefficients as well as η in a single optimization problem is known as the MKL problem. EasyMKL 3 is an MKL algorithm that estimates the relative weighting of each kernel by solving a quadratic optimization problem. Kernels that are not helpful for classification are down-weighted. EasyMKL's empirical effectiveness has been demonstrated across a large range of kernel numbers [bib_ref] EasyMKL: a scalable multiple kernel learning algorithm, Aiolli [/bib_ref] [bib_ref] A multimodal multiple kernel learning approach to Alzheimer's disease detection, Donini [/bib_ref]. All features were Z-scored across participants before projecting into kernels, separately for the training set and test set. We also employed L1 (min-max) normalization of kernels to project all their elements to the interval [0 1]. To ensure that any differences between combination methods did not reflect details of the classifier, the same EasyMKL algorithm was used in all cases (even when only N = 1 kernel in the case of Early combination).
The EasyMKL algorithm uses a regularization parameter, lambda ( ). A value of close to 1 penalizes less informative data, though potentially under-fits data, while a value of close to zero does not penalize less informative data, so potentially over-fits data (in that results may be 3 For more theoretical view including mathematical formulation of EasyMKL see [bib_ref] EasyMKL: a scalable multiple kernel learning algorithm, Aiolli [/bib_ref] Section 4 . sensitive to outliers in the training data;. Here, we used (stacked) nested cross-validation [bib_ref] Assessing and tuning brain decoders: cross-validation, caveats, and guidelines, Varoquaux [/bib_ref] by grid search with the same scoring as used for evaluation of the model performance to optimize , as implemented in the mkl_ens_nestedcv.m script in the GitHub repository. This function tunes for both the first stage of feature kernel combination and the second stage of decision combination. The value of for all analyses are reported in Supplementary Section S8.
Classification performance was estimated using 5-fold crossvalidation. Note this applied to both stages of Late combination, i.e., performance was always assessed using the untrained fold. Though the overall sample was unbalanced (with more HEC cases than MCI cases), the training set was always selected to be balanced, and the excess HEC participants were assigned to the test set. Given this imbalance in the test set, we report "balanced " classification accuracy, i.e., the mean accuracy across each class separately. Noise simulations in the Supplementary Section S1 confirmed that our estimation procedure was unbiased.
Cross-validation was repeated 1000 times with random selections of the data, in order to estimate classification reliability and to attenuate bias due to the actual order of the data which is recommended by [bib_ref] Assessing and tuning brain decoders: cross-validation, caveats, and guidelines, Varoquaux [/bib_ref]. Note that the specific random assignment of participants to training/test sets was matched when comparing different combination methods, meaning that classification accuracies could be directly subtracted for each comparison of interest, such that we could determine the percentage of the distribution of differences in classification accuracies that was greater than zero. This provides an approximation of the reliability of any improvement offered by one approach versus another (e.g., combined MEG and MRI versus MRI alone, or Intermediate vs Late combination of MEG and MRI).
## Analyses
Firstly, we evaluate classification performance of each modality i.e., MRI and MEG, separately, as well as from the confounding variables. We then combine (via Late combination) the confounding variables with either MRI or MEG, to test whether either imaging modality allows a reliable improvement relative to confounding variables alone. Secondly, we compare Early, Intermediate and Late combination of MRI and MEG modalities. Note that we start with one of MEG features i.e., the covariance of gradiometers in low gamma, but go on to consider other MEG features.
In supporting analyses, Section S1 of Supplementary material uses simulations to compare performance with different sets of signal and noise kernels, to evaluate how well the MKL approach works. In Supplementary Section S2, we repeat the first analyses in the main text for Early and Intermediate combinations, for completeness. In Supplementary Section S3, we try PCA as one of the common feature reduction methods, to see whether the classifier can take advantage of dimensionality reduction. In Supplementary Section S4, we show the classification accuracies for different types of MEG features alone, as well as a comparison of gradiometers versus magnetometers, and covariance versus variance. In Supplementary Section S5, we replace SVMs in the first stage of our Late combination with K-nearest neighbors, random forest, or multi-layer neural networks. In Supplementary Section S6, we use gray-matter volume in all voxels of the MRI image (instead of ROI data) for combining with MEG. Finally, in Supplementary Section S7, we combine MRI and MEG features for classification of the subset of MCI converters versus non-converters.
# Results
## Confounds and single mri and meg modalities
Panel a1 in [fig_ref] Figure 2: Left column [/fig_ref] shows the distribution across 1000 permutations of classification accuracies based on 8 kernels, each representing one of the potentially-confounding variables (COFs). The mean accuracy was 64.5%, and above chance (50%) on 100% of occasions. These results come from Late combination of the 8 kernels; results using Intermediate and Early combination are shown in Supplementary Section S2. This demonstrates that the two groups (MCI and Controls) were not perfectly matched in these potential confounds, but the reliable above-chance classification (which may be accidental or causal) provides a baseline to compare with classification using the MEG and MRI features.
For a single kernel based on the MRI features (of GM volume within 110 anatomical ROIs), Panel b1 of [fig_ref] Figure 2: Left column [/fig_ref] shows a mean accuracy of 71.4% (note that there is no distinction between Early, Intermediate and Late combination for a single kernel). To test whether this is a reliable improvement relative to COFs alone, Panel a2 of [fig_ref] Figure 2: Left column [/fig_ref] shows the distribution of differences between accuracies based on MRI versus COFs (when using the same permutations), which showed that MRI was more accurate on 95.5% of occasions. This demonstrates that MRI provides more information about MCI status than the potential other confounds considered.
For the MEG features, we start with the covariance across GRDs in the low Gamma range (see later for results using other MEG features). Panel c1 of [fig_ref] Figure 2: Left column [/fig_ref] shows a mean accuracy of 68.4%. While this MEG performance was higher than the baseline provided by the COFs on 86.6% of occasions (Panel b2), it was lower than for MRI on 78.5% of occasions (Panel c2). This finding that MRI is generally better than MEG is not surprising, since an MRI is also often used to define the MCI label, i.e., is likely to be biased (see Discussion). The more interesting comparison is whether combining MEG and MRI improves classification relative to MRI alone, which we return to after the next section.
## Adjusting for confounds
Panels d1 and e1 show results from Late combination of the COF kernels and either the MRI or MEG kernel, respectively. The addition of the COF kernels improves accuracy for both neuroimaging variables, but the more important result is in Panels d2 and e2, which show that these combinations are better than COFs alone on 99.7% and 90.7% of occasions for MRI and MEG respectively. This method of combining predictions from confounds (covariates) with those from features of interest (e.g. MRI) is arguably a better way to adjust for confounds than projecting them out of the features of interest themselves [bib_ref] Controlling for effects of confounding variables on machine learning predictions, Dinga [/bib_ref] [bib_ref] How to control for confounds in decoding analyses of neuroimaging data, Snoek [/bib_ref].shows classification accuracies when combining the MRI and MEG features at either early, intermediate or late stages. Panel a1 shows that accuracy for early combination is 69.8% which is less than MRI alone (71.4%). This demonstrates that concatenating features from different modalities is not an efficient way to combine them. For Intermediate combination, on the other hand, performance is improved on 87.4% of occasions relative to Early combination (Panel a2), with a mean accuracy of 74.3% (Panel b1). This demonstrates that combining feature kernels is better than concatenating features for these data. Late combination improves performance still further, improving on Intermediate combination on 90.6% of occasions (Panel b2), with a mean accuracy of 77.2% (Panel c1). This demonstrates that combining decision kernels is better than combining feature kernels for these data.
## Advantages of early, intermediate and late combination of mri and meg
The most important result is shown in Panel c2, which shows that Late combination of MEG and MRI improves classification accuracy compared to MRI alone on nearly 97.5% of occasions. This suggests that MEG provides information about MCI status that is complementary to that in MRI. A similar improvement occurred on 86.8% of occasions when the 8 COF kernels were also added [fig_ref] Figure 1: Schematic showing the three different combination stages used here [/fig_ref] , showing that this complementary information is also different from anything captured by the potential confounds. The same general improvement when adding MEG to MRI was also found when massively reducing the ratio of MEG features relative to MRI features by using PCA for feature selection (Supplementary Section S4), or dramatically increasing the number of MRI features by using voxels rather than ROIs (Supplementary Section S6). Similar improvements of multimodal integration were also found when using some popular classifiers other than SVMs (viz. KNN, random forest or neural networks), particularly for those classifiers that did better overall (Supplementary Section S5).
## Combining mri and other meg features
So far, we selected the COV of GRD in low Gamma as the MEG features, but it is possible that the above results are biased by this selection: i.e., when examining all 24 possibilities of variance/covariance (VAR/COV), gradiometers/magnetometers (GRD/MAG) and the 6 frequency bands, it is possible that one or more of them would show better performance when combined with MRI (than MRI alone) due to chance alone. We, therefore, repeated the above analyses across all frequency bands using both VAR and COV of both GRD and MAG. The results are shown in [fig_ref] Table 2: Exploring the MEG feature space [/fig_ref] , where the top numbers are the mean and standard deviation of classification performance (using Late combination of that MEG feature with MRI), the middle number shows the percentage of permutations in which this combination was better than MRI alone, and the bottom number shows the mean difference between MEG-MRI combination and MRI alone (see Supplementary Section S3 for raw performances for MEG alone).
Note that for low frequencies (Delta, Theta and Alpha), accuracy when combining them with MRI did not consistently exceed that for MRI alone (71.4%) -i.e. was better on around 50% occasions, as would be expected by chance. In fact, adding MEG sometimes reduced classification accuracy, which could either reflect estimation noise, or the fact that the MKL algorithm is not perfect at ignoring kernels that do not improve classification (see simulations in Supplementary Section S1).
For the Beta band, combined accuracy increased slightly, being better than MRI alone on 84% of occasions for the COV of GRD. For low Gamma, multimodal accuracy was consistently better for all four types of feature, with COV doing better than VAR/COV, and GRD doing better than MAG (see Supplementary Section S3 for more details). The same was also true of high Gamma. The overall finding that multimodal combination improved accuracy for over a third of the feature space suggests that the improvement is not a fluke occurrence.
# Discussion
The main finding of the present study was that certain features of MEG resting-state data, particularly in the low and high Gamma frequency range, improve the classification of individuals with MCI versus healthy controls when combined with features from structural MRI data. While classification accuracy with MEG alone never exceeded that for MRI alone, this is not necessarily surprising, since an MRI is typically used by the clinicians to support the diagnosis of MCI (i.e., giving MRI an unfair advantage). The important result was that combining MEG with MRI improved classification relative to MRI alone. This indicates that MEG contains complementary information about MCI. This information might include changes in functional activity and/or connectivity that precede structural change, at least as measured by regional gray-matter volume as here [bib_ref] Preclinical Alzheimer's disease: definition, natural history, and diagnostic criteria, Dubois [/bib_ref] [bib_ref] Beta amyloid, tau, neuroimaging, and cognition: sequence modeling of biomarkers for Alzheimer's..., Han [/bib_ref] [bib_ref] Age-specific and sex-specific prevalence of cerebral -amyloidosis, tauopathy, and neurodegeneration in cognitively..., Jack [/bib_ref] [bib_ref] Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic..., Jack [/bib_ref].
The second main finding of the present study was that Late combination of MEG and MRI data was best for multimodal classification, i.e., better than Intermediate or Early combination. Late combination here refers to combining the class predictions of classifiers trained on each modality separately, analogous to ensemble learning. Permutation tests showed that this combination at the "decision-level " reliably improved accuracy relative to Intermediate combination at the "feature-level ", where kernels derived from the features of each modality were combined directly. As expected, Intermediate combination via kernels was in turn better than simply concatenating (normalized) features into a single kernel. Note that these findings were obtained when the same classification algorithm (EasyMKL; [bib_ref] EasyMKL: a scalable multiple kernel learning algorithm, Aiolli [/bib_ref] -i.e., multi-kernel learning of support vector machines -was used for each level of multimodal combination.
A third finding was that the covariance (COV) of the planar gradiometers (GRD) generally provided the best MEG features, and it was important to consider high-frequency components of the MEG data, specifically the low (30-48 Hz) or high (52-86 Hz) Gamma range. Accuracy when using the Beta range did not produce such a large improvement over MRI alone, while that for Alpha, Theta or Delta provided no improvement. This suggests that the important information for MCI classification exists in frequencies above approximately 30 Hz (see below for further discussion).
A fourth outcome was the demonstration that potential confounds in any classification problem can be accommodated within a multimodal approach, in which confounds are combined with the features of interest during classification (which applies whether the combination is Early, Intermediate or Late). To "control for " such confounds, one can show that classification accuracy with the combined data reliably exceeds that for the confounds alone. This is better than the common approach of first adjusting the features of interest by the confounds (e.g., by projecting out of the features of interest anything that can be explained by a linear combination of the confounds, i.e., before creating the feature kernels), because it takes into account the potentially shared dependency between the features of interest and the confounds in their ability to predict the class [bib_ref] Controlling for effects of confounding variables on machine learning predictions, Dinga [/bib_ref]. Furthermore, some classifiers are non-linear, and therefore potentially sensitive to effects of confounds that cannot be removed from the features by linear methods.
## Complementary information in meg/eeg for mci classification
Previous studies have shown that EEG and/or MEG provide complementary information beyond structural MRI, whether that be in predicting age , or classifying types of dementia [bib_ref] Multimodal EEG-MRI in the differential diagnosis of Alzheimer's disease and dementia with..., Colloby [/bib_ref] [bib_ref] EEG and MRI data fusion for early diagnosis of Alzheimer's disease, Patel [/bib_ref] [bib_ref] Multimodal EEG, MRI and PET data fusion for Alzheimer's disease diagnosis, Polikar [/bib_ref]. Some of these have used evoked EEG responses during tasks (e.g., auditory oddballs [bib_ref] EEG and MRI data fusion for early diagnosis of Alzheimer's disease, Patel [/bib_ref] [bib_ref] Multimodal EEG, MRI and PET data fusion for Alzheimer's disease diagnosis, Polikar [/bib_ref] , which may provide further information on neurodegeneration than the resting-state data used here, though resting-state data are more common and easier to obtain, particularly in patients who might struggle with some tasks. In the data pa-per describing the BioFIND dataset, we reported validation analyses that showed that MEG power across all sensors, or power across all cortical sources, or connectivity between all pairs of sources (based the correlation of the power envelopes) achieved similar classifcation accuracies between 63 and 67%, comparable to the figure of 68% here for MEG alone (at least for COV of GRD in low Gamma). However, in that paper, we did not compare MEG classification with that from MRI. In a previous paper describing the smaller BioFIND project [bib_ref] Biomagnetic biomarkers for dementia: a pilot multicentre study with a recommended methodological..., Hughes [/bib_ref] , we did report preliminary findings that Intermediate combination of MEG and sMRI improves classification, using a subset of roughly half ( N = 168) of the present sample (that was available at that time). However, this finding used different features (interpolated 3D scalp-frequency power images for MEG and voxel-level GM images for sMRI) and did not establish the reliability of the improvement using permutation. The present work confirms the reliability of these findings and extends the approach to demonstrate the added value of Late combination of modalities, and the potential value of covariance, rather than power, across MEG sensors within certain frequency bands.
It is important to note that we have only used one type of structural brain information -namely gray-matter volume as estimated from a T1weighted MRI. Other MRI modalities (e.g., T2-weighted MRI, diffusionweighted MRI, magnetic resonance spectroscopy, MRS) -or indeed even other features from the current T1-weighted images, such as cortical thickness -might enable better MCI classification, to the extent that MEG no longer adds further improvement. Furthermore, other imaging techniques like PET might do better still, given their ability to measure neurotransmitters or molecular pathologies directly related to AD. However, our main purpose here was to compare MEG with the most common type of brain image available on individuals with MCI, and the most common type of informal inspection done by clinicians, i.e., looking for gray-matter atrophy.
We cannot tell whether the complementary information provided by MEG here relates to the fact that it measures brain function rather than brain structure, or that it provides a measurement of brain function with different spatial and temporal properties than other functional techniques like fMRI. Indeed, it would be interesting to apply the present multimodal classification approach to fMRI data, to see if MEG continues to provide more information than fMRI. If MEG does not improve classification beyond fMRI, then the key additional information for better MCI classification might simply be the inclusion of measures of brain function rather than structure; alternatively, if MEG does improve beyond fMRI, it may be that MEG captures neural activity more directly (bypassing the vascular confounds in fMRI) and/or neural activity that is beyond the temporal resolution of fMRI. Unfortunately, resting-state fMRI is not available in the BioFIND dataset, but is likely to be available together with MEG/EEG and sMRI in future cohorts being studied around the world. Likewise, one could test whether the improved spatial resolution of MEG over EEG also provides additional information for MCI classification and whether task-based MEG/EEG provides additional information beyond the resting-state data used here.
Finally, we cannot tell whether MEG provides complementary information beyond neuropsychological tests. The only cognitive test available on all participants in the current dataset is the MMSE, which is a very brief screening instrument. We did not compare classification with MEG to that with MMSE because the latter was one of the main determinants of the MCI diagnosis (even more so than MRI). It is possible that other more detailed cognitive assessments (not used in patient diagnosis) would be as, or more, effective than brain measures from MEG or MRI in detecting AD and predicting long-term outcome.
## Multimodal classification approaches
Our best classification accuracy of 77% when using MEG and MRI may not seem particularly impressive, for example, relative to figures reported in other papers using different modalities and datasets. We did explore some other common classifiers (such as KNN, random forest and multi-layer neural networks), which gave similar results (Supplementary Section S5). However, it is important to note that our aim was not simply to achieve the best classification possible. For example, while we found similar results after PCA to reduce the feature dimensionality (Supplementary Section S4), we could have employed more sophisticated feature selection approaches that might have improved classification accuracy, particularly given the large number of MEG features relative to participants, or we could have tried to minimize effects of field spread on our second-order (covariance) features by employing spatial filtering or Riemannian Embedding . Furthermore, we could have used neuroscientific knowledge to select features, for example, based on the knowledge that the medial temporal lobes include some of the structures affected in the earliest stages of AD [bib_ref] The clinical use of structural MRI in Alzheimer disease, Frisoni [/bib_ref] [bib_ref] Hippocampal volume and asymmetry in mild cognitive impairment and Alzheimer's disease: meta-analyses..., Shi [/bib_ref]. Rather, our aim was only to compare the relative performance of different modalities and different methods of combining those modalities, while holding other factors constant.
It is also important to note that the "MCI " label typically captures a range of aetiologies, of which neurodegeneration, and specifically AD, is only one. For example, some of the participants labeled as MCI in the BioFIND dataset may in fact have healthy brain structure and/or function (i.e., no evidence of early AD), but just perform poorly on cognitive tests because of other reasons like depression. Indeed, an appreciable proportion of clinically diagnosed MCI cases later turn out to have no detectable AD pathology [bib_ref] Early diagnosis of Alzheimer's disease: is MCI too late?, Petersen [/bib_ref]. Conversely, some participants labeled as healthy controls may have had early AD and impairments of brain function and/or structure, but performed normally on cognitive tests because of high pre-morbid ability or some form of "cognitive reserve " . This would mean it is difficult for any classifier to identify a consistent set of features (in this heterogeneous group) that perfectly predicts our two classes. These issues can only be resolved by longitudinal follow-up, possibly with additional biomarkers (e.g., CSF Tau levels) and ultimately post mortem examination to confirm who had AD. While we did analyze the follow-up data that are available on a subset of the participants labeled MCI in BioFIND (see Supplementary Section S7), the numbers are relatively small ( ∼50 converters and ∼50 non-converters), and further follow-up is needed, as well as additions from other datasets with larger numbers of cases (that are accessible to researchers, like BioFIND).
It is important to note that, while the BioFIND dataset may be the largest sample of individuals with MCI and controls with MEG data, it is still small for machine-learning approaches, relative to the potential number of MEG features. This dearth of training data may also explain why higher classification accuracies have been reported for other neuroimaging markers (e.g., sMRI) for which larger databases exist, such as ADNI ( http://adni.loni.usc.edu/about/ ). As a reference, using MRI alone on the ADNI database, reported an accuracy of 76% for classifying MCI cases versus healthy controls, using sophisticated deep neural net classifiers. Interestingly, this is only ∼4% more than achieved here using standard SVMs on a smaller set of MRI cases. However, the convolutional neural network architecture described can be considered for future studies focusing on the best classification performance where for example, sMRI features can be extracted using convolutional layers.
Furthermore, in situations with many more features than cases, overfitting is likely [bib_ref] Preventing over-fitting during model selection via Bayesian regularisation of the hyper-parameters, Cawley [/bib_ref] [bib_ref] Overcome support vector machine diagnosis overfitting, Han [/bib_ref]. This is a situation where Late combination might increase generalization to new datasets, by virtue of the combination of decisions being more robust to over-fitting [bib_ref] Stacked generalization, Wolpert [/bib_ref]. Indeed, the simulations in Panel d of Supplementary Section S1 confirm that Late combination can be better than Intermediate combination when more noise features are added, i.e., is more robust against the addition of weak (in terms of accuracy) classifiers trained on noisy features. Note however that Late combination is not always better than Intermediate combination, and multi-kernel combination is not always better than simple feature concatenation (Early combination), as can be seen for the 8 COFs in Supplementary Section S2, i.e., in situations with relatively low numbers of features.
## Optimal meg features for mci classification
Our finding that low and high Gamma frequencies provide the information that is complementary to MRI is consistent with some previous M/EEG studies of MCI or genetic risk that highlight the importance of the gamma band [bib_ref] Functional connectivity in mild cognitive impairment during a memory task: implications for..., Bajo [/bib_ref] [bib_ref] MEG detects abnormal hippocampal activity in amyloid-positive MCI, Luppi [/bib_ref] [bib_ref] Early disturbances of gamma band dynamics in mild cognitive impairment, Missonnier [/bib_ref] [bib_ref] Increased EEG gamma band activity in Alzheimer's disease and mild cognitive impairment, Deursen [/bib_ref]. However, differences (e.g., in source level power and/or connectivity) were more often found at lower frequencies. For example, [bib_ref] Biomagnetic biomarkers for dementia: a pilot multicentre study with a recommended methodological..., Hughes [/bib_ref] , [bib_ref] MEG spectral analysis in subtypes of mild cognitive impairment, López [/bib_ref] , [bib_ref] The role of magnetoencephalography in the early stages of alzheimer's disease, López-Sanz [/bib_ref] , [bib_ref] The importance of the validation of M/EEG with current biomarkers in Alzheimer's..., Maestú [/bib_ref] , [bib_ref] Electromagnetic signatures of the preclinical and prodromal stages of Alzheimer's disease, Nakamura [/bib_ref] have argued that the alpha band is best for distinguishing MCI versus controls. One possibility is that the information about MCI status provided by Alpha power is correlated with the gray-matter atrophy provided by MRI, which is why we did not find any improvement when combining MRI with MEG covariance in this frequency band. However, it is worth noting that the best classification accuracy with Alpha alone ( ∼59%) was still considerably lower than for Gamma alone ( ∼68%; Supplementary Section S3). Another reason for this discrepancy in the literature may reflect the features used, for example, the amplitude or frequency of the prominent Alpha peak in MEG and EEG power spectra is a type of feature that is not simply (e.g. linearly) derivable from the Alpha covariance matrix used here. Other discrepancies may owe to the use of different and relatively small samples ( N < 100), possibly with different definitions of individuals with MCI, and to the wide range of methodological approaches [bib_ref] M/EEG-based bio-markers to predict the MCI and Alzheimer's disease: a review from..., Yang [/bib_ref]. Indeed, when we examined the classification of converters versus non-converters within the MCI group only, we found a trend for lower frequencies like Delta and Theta to provide more complementary information than the Gamma that was optimal for classifying the MCI versus Control groups (Supplementary Section S7). Thus, we do not wish to claim that Gamma frequencies, or more specifically covariance of planar gradiometers, are always the best MEG features to use for early detection of AD. We only used the sensor covariance matrix as a simple but inclusive measure of functional activity and connectivity and explored the range of frequency bands because of prior evidence that frequency matters. It is worth mentioning that there was some evidence that covariance classified better than variance (Supplementary Section S3), which may be because the sensor covariance, though not a pure measure of brain connectivity, might capture more aspects of connectivity than sensor variance. Nonetheless, future studies could use the present framework to ask whether specific MEG features offer additional information about MCI over other MEG features.
## Code availability
The custom written codes to implement all validation analyses is available on GitHub ( https://github.com/delshadv/MRI _ MEG _ Combination ). All MEG and MRI features as well as other derived variables are available in comma-separated value (.csv) files in the "derived " directory within the repository. The raw data are available on the DPUK website ( https://portal.dementiasplatform.uk/Apply ) cited in the main paper.
# Author statement
We freely provide the codes in the GitHub repository i.e. https: //github.com/delshadv for the kernel-based approach, as well as for the MRI and MEG pre-processing and feature extraction steps while the raw data are available on request from the Dementia Platform UK (DPUK). Necessary links to access data are available in the main text. Furthermore, all features that are necessary to reproduce main results like MEG variance, MEG covariance and MRI ROI data are available on both paper's GitHub and DPUK portal in case one does not have the interest or time to reproduce the preprocessing steps.
## Declaration of competing interest
None declared.
## Credit authorship contribution statement
[fig] Figure 1: Schematic showing the three different combination stages used here: Early (feature concatenation), Intermediate (kernel combination) and Late (decision combination). N is number of kernels. GM = Gray-Matter, HOA = Harvard-Oxford Atlas, COFs = (potential) Confounds. [/fig]
[fig] Figure 2: Left column: Classification accuracies (chance = 50%) from 1000 random permutations using Late combinations of MRI, MEG (covariance of gradiometers in low-gamma band) and the 8 potential confounding variables. Right column: Differences in classification performance for each permutation when comparing various combinations of features in left column (where 0 = means no difference). "A,B " means combining two (or nine -in presence of confounds) predictions derived from models trained using modality-type A and modality-type B. [/fig]
[fig] Figure 3: Left column: Classification accuracies (chance = 50%) from 1000 random permutations using MEG, MRI with Early, Intermediate and Late combinations (see methods). Right column: Differences in classification performance for each permutation when comparing various combinations approach (where 0 = means no difference). [/fig]
[table] Table 1: Characteristics of BioFIND participants with clean MRIs. CBU = Cognition & Brain Sciences Unit (Cambridge); CTB = center for Biomedical Technology (Madrid). Translation calculated every second from Maxfilter stage (seeVaghari et al. 2021 ). Distance from sensors was calculated after coregistering the MEG to the MRI, and averaging the Euclidean distance between every sensor and each of 2562 vertices on a cortical mesh. SD = standard deviation. MMSE = Mini-Mental State Examination. [/table]
[table] Table 2: Exploring the MEG feature space. The top numbers show mean (and SD in brackets) of classification accuracy when combining the relevant MEG feature with MRI; the middle number shows the percentage of permutations in which this accuracy exceeded that of MRI alone (71.4%), where chance = 50%; the bottom number shows the average difference between MEG-MRI combination and MRI alone. [/table]
[table] Delshad Vaghari: Conceptualization, Formal analysis, Writingoriginal draft. Ehsanollah Kabir: Writing -original draft. Richard N. Henson: Conceptualization, Writing -original draft. [/table]
|
Prevalence of Human Papilloma Virus Infection in Bladder Cancer: A Systematic Review
# Introduction
Bladder cancer (BCa) is the 10th most prevalent cancer globally, with >572,000 and >212,000 incident cases and deaths. The highest rates of BCa are registered in Southern and Western Europe and North America, with a higher incidence in men [bib_ref] Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for..., Sung [/bib_ref]. Several risk factors were found; most of them are related to personal behavior (i.e., diet and smoking), socioeconomic status (i.e., the accessibility to health services and delay in diagnosis), and environmental and occupational exposure to chemical substances or infectious diseases [bib_ref] Bladder cancer: Race differences in extent of disease at diagnosis, Prout [/bib_ref]. Moreover, epidemiological studies revealed a higher risk in men and a worse prognosis in black males [bib_ref] Metabolomic Approaches for Detection and Identification of Biomarkers and Altered Pathways in..., Di Meo [/bib_ref]. BCa histological subtypes can vary: urothelial BCa, previously classified as transitional cell carcinoma (TCC) and the predominant histological type, accounts for~90% and is mainly related to chemical exposure, whereas squamous cell carcinoma (SCC, 5%) is associated with chronic inflammation and persistent infections (Schistosoma spp. in Africa) [bib_ref] Epidemiology of Bladder Cancer, Saginala [/bib_ref]. Adenocarcinoma (2%), sarcoma, and small cell carcinoma are less incident forms.
While the role played by human papillomavirus (HPV) in the development of cervical, anogenital, and oropharyngeal cancers was proven [bib_ref] Distribution of HPV Genotypes in Patients with a Diagnosis of Anal Cancer..., Muresu [/bib_ref] [bib_ref] Italian observational study on HPV infection, E6, and p16 expression in men..., Muresu [/bib_ref] [bib_ref] Worldwide trend in human papillomavirus-attributable cancer incidence rates between 1990 and 2012..., Wu [/bib_ref] , the causative relationship between HPV and BCa still remains controversial, with a high variable prevalence attributed to the study design, the enrolled population, and the HPV detection methods [bib_ref] Human papillomavirus infection and bladder cancer risk: A meta-analysis, Li [/bib_ref]. Understanding the role of HPV in BCa could have relevant diagnostic, therapeutic, and preventive implications. The current systematic review is aimed to assess the prevalence of HPV infection in BCa, focusing on patients' clinical and epidemiological characteristics.
# Materials and methods
## Search strategy
A systematic literature review aimed at retrieving papers focused on the prevalence of HPV infection in BCa was carried out from its inception to 31st December 2021. The literature search was performed using PubMed and Scopus, selecting the key words "Human Papillomavirus", "HPV", or "Papillomavirus" and "bladder cancer" or "bladder carcinoma", combined in different strings. No restrictions related to age of patients, setting, or time of the study were chosen.
Lists of references of all selected articles were screened to find other eligible studies not included in the above-mentioned databases.
## Study selection and inclusion criteria
Case-control, cross-sectional, and cohort studies reporting HPV infection prevalence were selected.
The following inclusion criteria were considered:
(1) Studies dealing with patients with primary BCa;
(2) Studies describing molecular and non-molecular HPV detection methods on fresh or FFPE (Formalin Fixed Paraffin Embedded) bladder biopsies; (3) Studies focused on the following medical conditions: SCC, urothelial carcinoma (UC), and transitional cell carcinoma (TCC);
Articles were excluded for the following reasons:
(1) Review articles, abstracts, letters, commentaries, correspondences, case-reports, and case-series enrolling <10 subjects; (2) Use of languages other than English;
(3) Secondary malignancies located in the bladder.
Article selection and data extraction were performed by two Authors and doublechecked (M.N. and D.B.), while discrepancies of opinions or disagreement were resolved by a third investigator (S.G.).
## Data extraction
Qualitative and quantitative variables were collected in an ad hoc electronic form. The following variables were collected: first author's last name; title of the article; year and country/countries of the study; period of the study; epidemiological study design; sample size; sex and age; type of samples; HPV detection methods; histological subtypes; tumor grading; HPV prevalence.
No ethical approval was needed given the anonymized and aggregated nature of the data.
## Study quality assessment
Inter-rater agreement was~100% for the phases of study selection and data extraction, and the few inconsistencies were resolved by consensus and with the support of a third investigator (G.S.).
Guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) were followed to guide the process of the systematic review [bib_ref] The PRISMA 2020 statement: An updated guideline for reporting systematic reviews, Page [/bib_ref].
The Newcastle-Ottawa Scalewas used to assess the quality of the included casecontrol studies by evaluating selection, comparability, and exposure criteria, through four, two, and three items, respectively. The Joanna Briggs Institute Critical Appraisal tools (JBI), applied for analytical cross-sectional studies where the control group is missing, consists of eight items aimed at evaluating the risk of bias: high, moderate, or low risk of bias was assigned when positive answers were ≤49%, between 50% and 75%, or >75%, respectively [bib_ref] Chapter 2: Systematic reviews of qualitative evidence, Lockwood [/bib_ref] (Table S1a,b).
# Statistical analysis
Qualitative and quantitative variables were summarized with absolute and relative (percentage) frequencies and means/medians [standard deviation (SD), interquartile range (IQR)] respectively.
Forest plots were used to show pooled risk differences of the selected outcomes and interval (95% confidence interval, CI) estimates, as well as the weight of the sample size of the recruited studies. The I2 indicator (low, medium, and high heterogeneity expressed as <25%, ≥25%-<50%, ≥50%, respectively) showed the association between true variability and overall variation.
Fixed and random-effects models were chosen depending on the estimated betweenstudy heterogeneity. A two-tailed p-value less than 0.05 was deemed statistically significant. The statistical software Stata version 17 (StataCorp, College Station, TX, USA) and StatsDirect version 3.1.12 (StatsDirect Ltd., Willar, UK).
# Results
## Study selection
A total of 637 articles were identified through electronic database searches; 162 (26.2%) were excluded for being duplicates, and then, a total of 475 studies were screened by titles and abstracts. Fifty-six (11.8%) full texts were evaluated, and ten (17.9%) were excluded for the following reasons: tissue samples did not include bladder tissues (n = 5), full text was not available (n = 2), case-report (n = 1), review (n = 1), and dataset described in another study (n = 1). A total of 46 (46/56; 82.1%) manuscripts were included in the review (Table 1; [fig_ref] Figure 1: PRISMA 2020 flow diagram for new systematic reviews, which included searches of... [/fig_ref].
## Quality assessment
Twelve (34.3%) cross-sectional studies were deemed at moderate risk of bias, whereas 23 (65.7%) were classified as low-risk . Nine (81.8%) cross-sectional studies were deemed to be medium-quality, whereas two (18.2%) were high-quality [fig_ref] Table 3: NOS quality assessment table [/fig_ref]. . JBI risk of bias assessment table. Eight items per study were evaluated and the risk of bias was calculated on the number of positive answers. y = yes, n = no, u = unclear. Moderate = where positive answers were between 50% and 75%; low = where positive answers were above 75%.
## Ref. first author
Were "* " : Each study was awarded one star per item. The score is the sum of the awarded stars, ranged from 0 to 9 (high-quality, >7 stars; medium-quality, 4-6 stars; poor-quality, <4 stars).
## Study characteristics
Studies were published during the period 1995 [bib_ref] Detection of human papillomavirus DNA sequences in cancer of the urinary bladder..., Gopalkrishna [/bib_ref] [bib_ref] Human papillomavirus DNA and abnormal p53 expression in carcinoma of the urinary..., Kamel [/bib_ref] [bib_ref] Analysis of p53 tumor suppressor gene mutations and human papillomavirus infection in..., Kim [/bib_ref] [bib_ref] Human papillomavirus in transitional cell carcinoma of the urinary bladder, Larue [/bib_ref] -2021 [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref]. Patients were enrolled between 1985 [bib_ref] Human papillomavirus is not an etiologic agent of urothelial inverted papillomas, Alexander [/bib_ref] and 2019 [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref]. The epidemiological study types were observational retrospective [bib_ref] Human papillomavirus and schistosomiasis associated bladder cancer, Cooper [/bib_ref] [bib_ref] Human papillomaviruses 6/11, 16/18 and 31/33/51 are not associated with squamous cell..., Westenend [/bib_ref].5%) [bib_ref] Koilocytosis: Correlations with high-risk HPV and its comparison on tissue sections and..., Aggarwal [/bib_ref] [bib_ref] p16 expression is not associated with human papillomavirus in urinary bladder squamous..., Alexander [/bib_ref] [bib_ref] Human papillomavirus is not an etiologic agent of urothelial inverted papillomas, Alexander [/bib_ref] [bib_ref] The expression patterns of p53 and p16 and an analysis of a..., Alexander [/bib_ref] [bib_ref] Prevalence of six types of human papillomavirus in inverted papilloma and papillary..., Chan [/bib_ref] [bib_ref] High-risk human papillomavirus DNA detected in primary squamous cell carcinoma of urinary..., Chapman-Fredricks [/bib_ref] [bib_ref] Prevalence of high-risk human papillomavirus in primary squamous cell carcinoma of urinary..., Collins [/bib_ref] [bib_ref] Human papillomavirus and schistosomiasis associated bladder cancer, Cooper [/bib_ref] [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Prevalence of papillomavirus, Epstein-Barr virus, cytomegalovirus, and herpes simplex virus type 2..., Gazzaniga [/bib_ref] [bib_ref] Loss of Cell Differentiation in HPV-Associated Bladder Cancer, Golovina [/bib_ref] [bib_ref] Detection of human papillomavirus DNA sequences in cancer of the urinary bladder..., Gopalkrishna [/bib_ref] [bib_ref] Human Papilloma Virus DNA in Tumor Tissue and Urine in Different Stage..., Javanmard [/bib_ref] [bib_ref] Human papillomavirus DNA and abnormal p53 expression in carcinoma of the urinary..., Kamel [/bib_ref] [bib_ref] Analysis of p53 tumor suppressor gene mutations and human papillomavirus infection in..., Kim [/bib_ref] [bib_ref] Human papillomavirus in transitional cell carcinoma of the urinary bladder, Larue [/bib_ref] [bib_ref] Defining the frequency of human papillomavirus and polyomavirus infection in urothelial bladder..., Llewellyn [/bib_ref] [bib_ref] Human papillomavirus DNA as a factor determining the survival of bladder cancer..., Lopez-Beltran [/bib_ref] [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] [bib_ref] Condyloma Acuminatum of Urinary Bladder: Relation to Squamous Cell Carcinoma, Samarska [/bib_ref] [bib_ref] Low frequency of human papillomavirus infection in initial papillary bladder tumors, Simoneau [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] [bib_ref] Human papillomaviruses 6/11, 16/18 and 31/33/51 are not associated with squamous cell..., Westenend [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref] [bib_ref] Role of human papillomavirus in the development of urothelial carcinoma, Yavuzer [/bib_ref] [bib_ref] Does human papillomavirus play a role in the development of bladder transitional..., Youshya [/bib_ref] , case-control [bib_ref] The PRISMA 2020 statement: An updated guideline for reporting systematic reviews, Page [/bib_ref] [bib_ref] Prevalence of six types of human papillomavirus in inverted papilloma and papillary..., Chan [/bib_ref].9%) [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Human papillomavirus and p16 expression in inverted papillomas of the urinary bladder, Gould [/bib_ref] [bib_ref] Detection of human papillomavirus infection and p16 immunohistochemistry expression in bladder cancer..., Kim [/bib_ref] [bib_ref] Human Papillomavirus in Urothelial Carcinoma of Bladder: An Indian study, Mete [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Is there a correlation between HPV and urinary bladder carcinoma?, Shaker [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Overexpression of p16(INK4a) in urothelial carcinoma in situ is a marker for..., Steinestel [/bib_ref] [bib_ref] Human papillomavirus associated with bladder carcinoma? Analysis by polymerase chain reaction, Tekin [/bib_ref] , prospective [bib_ref] Human papillomavirus is not an etiologic agent of urothelial inverted papillomas, Alexander [/bib_ref].4%) [bib_ref] Investigation of human papillomavirus in bladder cancer in a series of Tunisian..., Ben Selma [/bib_ref] [bib_ref] Human papillomavirus detection in Moroccan patients with bladder cancer, Berrada [/bib_ref] [bib_ref] Human papilloma virus and p53 expression in bladder cancer in Egypt: Relationship..., Tel [/bib_ref] [bib_ref] Human papillomavirus infection and transitional cell carcinoma of the bladder. Immunohistochemistry and..., López-Beltrán [/bib_ref] [bib_ref] Association of human papilloma virus (HPV) infection with oncological outcomes in urothelial..., Moghadam Ohadian [/bib_ref] [bib_ref] Low prevalence of HPV detection and genotyping in non-muscle invasive bladder cancer..., Pichler [/bib_ref] [bib_ref] Human papilloma virus is not detectable in samples of urothelial bladder cancer..., Schmid [/bib_ref] [bib_ref] Etiological correlation of human papillomavirus infection in the development of female bladder..., Shigehara [/bib_ref] , and comparative (1, 2.2%) [bib_ref] Urothelial bladder carcinoma and viral infections: Different association with human polyomaviruses and..., Fioriti [/bib_ref]. Most of them were single-center (44/46, 95.7%) [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] Koilocytosis: Correlations with high-risk HPV and its comparison on tissue sections and..., Aggarwal [/bib_ref] [bib_ref] p16 expression is not associated with human papillomavirus in urinary bladder squamous..., Alexander [/bib_ref] , and only two (4.3%) were multi-center [bib_ref] Human papillomavirus is not an etiologic agent of urothelial inverted papillomas, Alexander [/bib_ref] [bib_ref] The expression patterns of p53 and p16 and an analysis of a..., Alexander [/bib_ref]. Single-center studies were performed in Europe [bib_ref] The expression patterns of p53 and p16 and an analysis of a..., Alexander [/bib_ref] [bib_ref] Human papillomavirus DNA as a factor determining the survival of bladder cancer..., Lopez-Beltran [/bib_ref].9%) [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Urothelial bladder carcinoma and viral infections: Different association with human polyomaviruses and..., Fioriti [/bib_ref] [bib_ref] Prevalence of papillomavirus, Epstein-Barr virus, cytomegalovirus, and herpes simplex virus type 2..., Gazzaniga [/bib_ref] [bib_ref] Loss of Cell Differentiation in HPV-Associated Bladder Cancer, Golovina [/bib_ref] [bib_ref] Human papillomavirus DNA and abnormal p53 expression in carcinoma of the urinary..., Kamel [/bib_ref] [bib_ref] Defining the frequency of human papillomavirus and polyomavirus infection in urothelial bladder..., Llewellyn [/bib_ref] [bib_ref] Human papillomavirus DNA as a factor determining the survival of bladder cancer..., Lopez-Beltran [/bib_ref] [bib_ref] Human papillomavirus infection and transitional cell carcinoma of the bladder. Immunohistochemistry and..., López-Beltrán [/bib_ref] [bib_ref] Low prevalence of HPV detection and genotyping in non-muscle invasive bladder cancer..., Pichler [/bib_ref] [bib_ref] Condyloma Acuminatum of Urinary Bladder: Relation to Squamous Cell Carcinoma, Samarska [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Human papilloma virus is not detectable in samples of urothelial bladder cancer..., Schmid [/bib_ref] [bib_ref] Overexpression of p16(INK4a) in urothelial carcinoma in situ is a marker for..., Steinestel [/bib_ref] [bib_ref] Human papillomavirus associated with bladder carcinoma? Analysis by polymerase chain reaction, Tekin [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] [bib_ref] Human papillomaviruses 6/11, 16/18 and 31/33/51 are not associated with squamous cell..., Westenend [/bib_ref] [bib_ref] Role of human papillomavirus in the development of urothelial carcinoma, Yavuzer [/bib_ref] [bib_ref] Does human papillomavirus play a role in the development of bladder transitional..., Youshya [/bib_ref] , Asia (14, 31.8%) [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] Koilocytosis: Correlations with high-risk HPV and its comparison on tissue sections and..., Aggarwal [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Prevalence of six types of human papillomavirus in inverted papilloma and papillary..., Chan [/bib_ref] [bib_ref] Detection of human papillomavirus DNA sequences in cancer of the urinary bladder..., Gopalkrishna [/bib_ref] [bib_ref] Human Papilloma Virus DNA in Tumor Tissue and Urine in Different Stage..., Javanmard [/bib_ref] [bib_ref] Analysis of p53 tumor suppressor gene mutations and human papillomavirus infection in..., Kim [/bib_ref] [bib_ref] Detection of human papillomavirus infection and p16 immunohistochemistry expression in bladder cancer..., Kim [/bib_ref] [bib_ref] Human Papillomavirus in Urothelial Carcinoma of Bladder: An Indian study, Mete [/bib_ref] [bib_ref] Association of human papilloma virus (HPV) infection with oncological outcomes in urothelial..., Moghadam Ohadian [/bib_ref] [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Etiological correlation of human papillomavirus infection in the development of female bladder..., Shigehara [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref] , America (6, 13.6%) [bib_ref] p16 expression is not associated with human papillomavirus in urinary bladder squamous..., Alexander [/bib_ref] [bib_ref] High-risk human papillomavirus DNA detected in primary squamous cell carcinoma of urinary..., Chapman-Fredricks [/bib_ref] [bib_ref] Prevalence of high-risk human papillomavirus in primary squamous cell carcinoma of urinary..., Collins [/bib_ref] [bib_ref] Human papillomavirus and p16 expression in inverted papillomas of the urinary bladder, Gould [/bib_ref] [bib_ref] Human papillomavirus in transitional cell carcinoma of the urinary bladder, Larue [/bib_ref] [bib_ref] Low frequency of human papillomavirus infection in initial papillary bladder tumors, Simoneau [/bib_ref] , and Africa (6, 13.6%) [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Investigation of human papillomavirus in bladder cancer in a series of Tunisian..., Ben Selma [/bib_ref] [bib_ref] Human papillomavirus detection in Moroccan patients with bladder cancer, Berrada [/bib_ref] [bib_ref] Human papillomavirus and schistosomiasis associated bladder cancer, Cooper [/bib_ref] [bib_ref] Human papilloma virus and p53 expression in bladder cancer in Egypt: Relationship..., Tel [/bib_ref] [bib_ref] Is there a correlation between HPV and urinary bladder carcinoma?, Shaker [/bib_ref].
## Characteristics of the study samples
The sample size ranged from 10 [bib_ref] Detection of human papillomavirus DNA sequences in cancer of the urinary bladder..., Gopalkrishna [/bib_ref] to 689 [bib_ref] Defining the frequency of human papillomavirus and polyomavirus infection in urothelial bladder..., Llewellyn [/bib_ref] patients, for a total of 3975 subjects. Information on gender was reported by 36 (73.5%) studies [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] Koilocytosis: Correlations with high-risk HPV and its comparison on tissue sections and..., Aggarwal [/bib_ref] [bib_ref] p16 expression is not associated with human papillomavirus in urinary bladder squamous..., Alexander [/bib_ref] [bib_ref] Human papillomavirus is not an etiologic agent of urothelial inverted papillomas, Alexander [/bib_ref] [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Investigation of human papillomavirus in bladder cancer in a series of Tunisian..., Ben Selma [/bib_ref] [bib_ref] Human papillomavirus detection in Moroccan patients with bladder cancer, Berrada [/bib_ref] [bib_ref] High-risk human papillomavirus DNA detected in primary squamous cell carcinoma of urinary..., Chapman-Fredricks [/bib_ref] [bib_ref] Prevalence of high-risk human papillomavirus in primary squamous cell carcinoma of urinary..., Collins [/bib_ref] [bib_ref] Human papillomavirus and schistosomiasis associated bladder cancer, Cooper [/bib_ref] [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Urothelial bladder carcinoma and viral infections: Different association with human polyomaviruses and..., Fioriti [/bib_ref] [bib_ref] Prevalence of papillomavirus, Epstein-Barr virus, cytomegalovirus, and herpes simplex virus type 2..., Gazzaniga [/bib_ref] [bib_ref] Loss of Cell Differentiation in HPV-Associated Bladder Cancer, Golovina [/bib_ref] [bib_ref] Detection of human papillomavirus DNA sequences in cancer of the urinary bladder..., Gopalkrishna [/bib_ref] [bib_ref] Human papillomavirus and p16 expression in inverted papillomas of the urinary bladder, Gould [/bib_ref] [bib_ref] Human papilloma virus and p53 expression in bladder cancer in Egypt: Relationship..., Tel [/bib_ref] [bib_ref] Human Papilloma Virus DNA in Tumor Tissue and Urine in Different Stage..., Javanmard [/bib_ref] [bib_ref] Detection of human papillomavirus infection and p16 immunohistochemistry expression in bladder cancer..., Kim [/bib_ref] [bib_ref] Human papillomavirus DNA as a factor determining the survival of bladder cancer..., Lopez-Beltran [/bib_ref] [bib_ref] Human papillomavirus infection and transitional cell carcinoma of the bladder. Immunohistochemistry and..., López-Beltrán [/bib_ref] [bib_ref] Human Papillomavirus in Urothelial Carcinoma of Bladder: An Indian study, Mete [/bib_ref] [bib_ref] Association of human papilloma virus (HPV) infection with oncological outcomes in urothelial..., Moghadam Ohadian [/bib_ref] [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] [bib_ref] Low prevalence of HPV detection and genotyping in non-muscle invasive bladder cancer..., Pichler [/bib_ref] [bib_ref] Condyloma Acuminatum of Urinary Bladder: Relation to Squamous Cell Carcinoma, Samarska [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Human papilloma virus is not detectable in samples of urothelial bladder cancer..., Schmid [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Etiological correlation of human papillomavirus infection in the development of female bladder..., Shigehara [/bib_ref] [bib_ref] Overexpression of p16(INK4a) in urothelial carcinoma in situ is a marker for..., Steinestel [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] [bib_ref] Human papillomaviruses 6/11, 16/18 and 31/33/51 are not associated with squamous cell..., Westenend [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref] [bib_ref] Role of human papillomavirus in the development of urothelial carcinoma, Yavuzer [/bib_ref] , including 555 and 1969 females and males, respectively. The mean/median age ranged from 47 [bib_ref] Human papillomavirus and schistosomiasis associated bladder cancer, Cooper [/bib_ref] to 74.8 [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] years .
The majority of the samples were FFPE (2706/3518; 76.9%) , and only 812 (23.1%) were fresh tissue specimens [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Urothelial bladder carcinoma and viral infections: Different association with human polyomaviruses and..., Fioriti [/bib_ref] [bib_ref] Prevalence of papillomavirus, Epstein-Barr virus, cytomegalovirus, and herpes simplex virus type 2..., Gazzaniga [/bib_ref] [bib_ref] Human Papillomavirus in Urothelial Carcinoma of Bladder: An Indian study, Mete [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Human papilloma virus is not detectable in samples of urothelial bladder cancer..., Schmid [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Human papillomavirus associated with bladder carcinoma? Analysis by polymerase chain reaction, Tekin [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref] [bib_ref] Does human papillomavirus play a role in the development of bladder transitional..., Youshya [/bib_ref]. Information on the type of specimen was not available for three (3/46; 6.5%) studies [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Defining the frequency of human papillomavirus and polyomavirus infection in urothelial bladder..., Llewellyn [/bib_ref] [bib_ref] Low frequency of human papillomavirus infection in initial papillary bladder tumors, Simoneau [/bib_ref].
Histological classification was available for 42/46 (91.3%): TCC were the most prevalent (1445/2792; 51.8%) type, followed by UC (1098/2792; 39.3%) and SCC (249/2792; 8.9%) [fig_ref] Table 3: NOS quality assessment table [/fig_ref].
Sixteen (34.8%) studies [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] Investigation of human papillomavirus in bladder cancer in a series of Tunisian..., Ben Selma [/bib_ref] [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Loss of Cell Differentiation in HPV-Associated Bladder Cancer, Golovina [/bib_ref] [bib_ref] Human papillomavirus DNA and abnormal p53 expression in carcinoma of the urinary..., Kamel [/bib_ref] [bib_ref] Human papillomavirus in transitional cell carcinoma of the urinary bladder, Larue [/bib_ref] [bib_ref] Human papillomavirus DNA as a factor determining the survival of bladder cancer..., Lopez-Beltran [/bib_ref] [bib_ref] Human papillomavirus infection and transitional cell carcinoma of the bladder. Immunohistochemistry and..., López-Beltrán [/bib_ref] [bib_ref] Low prevalence of HPV detection and genotyping in non-muscle invasive bladder cancer..., Pichler [/bib_ref] [bib_ref] Human papilloma virus is not detectable in samples of urothelial bladder cancer..., Schmid [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Etiological correlation of human papillomavirus infection in the development of female bladder..., Shigehara [/bib_ref] [bib_ref] Low frequency of human papillomavirus infection in initial papillary bladder tumors, Simoneau [/bib_ref] [bib_ref] Human papillomavirus associated with bladder carcinoma? Analysis by polymerase chain reaction, Tekin [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] , for a total of 1428 samples, reported the grading, following the recommendations of the American Joint Committee on Cancer [AJCC Cancer Staging Manual. 7th ed. New York, NY: Springer; 2010.]. A total of 580 (40.6%), 513 (35.9%), and 335 (23.5%) tumors were classified as moderate (G2), poor (G3), and well (G1) differentiated, respectively. 1049 specimens [bib_ref] Koilocytosis: Correlations with high-risk HPV and its comparison on tissue sections and..., Aggarwal [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Human papillomavirus detection in Moroccan patients with bladder cancer, Berrada [/bib_ref] [bib_ref] Human papilloma virus and p53 expression in bladder cancer in Egypt: Relationship..., Tel [/bib_ref] [bib_ref] Human Papilloma Virus DNA in Tumor Tissue and Urine in Different Stage..., Javanmard [/bib_ref] [bib_ref] Analysis of p53 tumor suppressor gene mutations and human papillomavirus infection in..., Kim [/bib_ref] [bib_ref] Detection of human papillomavirus infection and p16 immunohistochemistry expression in bladder cancer..., Kim [/bib_ref] [bib_ref] Human Papillomavirus in Urothelial Carcinoma of Bladder: An Indian study, Mete [/bib_ref] [bib_ref] Association of human papilloma virus (HPV) infection with oncological outcomes in urothelial..., Moghadam Ohadian [/bib_ref] [bib_ref] Low prevalence of HPV detection and genotyping in non-muscle invasive bladder cancer..., Pichler [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Is there a correlation between HPV and urinary bladder carcinoma?, Shaker [/bib_ref] [bib_ref] Etiological correlation of human papillomavirus infection in the development of female bladder..., Shigehara [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref] [bib_ref] Role of human papillomavirus in the development of urothelial carcinoma, Yavuzer [/bib_ref] were classified according to the guidelines of the European Association of Urology [bib_ref] Prognostic Performance and Reproducibility of the 1973 and 2004/2016 World Health Organization..., Soukup [/bib_ref] , with 541 (51.6%) low-and 508 (48.4%) high-grade lesions [fig_ref] Table 3: NOS quality assessment table [/fig_ref].
The most frequent HPV detection method was molecular (38/46; 82.6%) [bib_ref] Koilocytosis: Correlations with high-risk HPV and its comparison on tissue sections and..., Aggarwal [/bib_ref] [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Investigation of human papillomavirus in bladder cancer in a series of Tunisian..., Ben Selma [/bib_ref] [bib_ref] Human papillomavirus detection in Moroccan patients with bladder cancer, Berrada [/bib_ref] [bib_ref] Prevalence of six types of human papillomavirus in inverted papilloma and papillary..., Chan [/bib_ref] [bib_ref] High-risk human papillomavirus DNA detected in primary squamous cell carcinoma of urinary..., Chapman-Fredricks [/bib_ref] [bib_ref] Human papillomavirus and schistosomiasis associated bladder cancer, Cooper [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Etiological correlation of human papillomavirus infection in the development of female bladder..., Shigehara [/bib_ref] [bib_ref] Low frequency of human papillomavirus infection in initial papillary bladder tumors, Simoneau [/bib_ref] [bib_ref] Overexpression of p16(INK4a) in urothelial carcinoma in situ is a marker for..., Steinestel [/bib_ref] [bib_ref] Human papillomavirus associated with bladder carcinoma? Analysis by polymerase chain reaction, Tekin [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref] [bib_ref] Role of human papillomavirus in the development of urothelial carcinoma, Yavuzer [/bib_ref] [bib_ref] Does human papillomavirus play a role in the development of bladder transitional..., Youshya [/bib_ref] , whereas a non-molecular technique (i.e., immunohistochemistry-IHC and/or in situ hybridization-ISH) was employed in 34 (73.9%) studies [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] p16 expression is not associated with human papillomavirus in urinary bladder squamous..., Alexander [/bib_ref] [bib_ref] Human papillomavirus is not an etiologic agent of urothelial inverted papillomas, Alexander [/bib_ref] [bib_ref] The expression patterns of p53 and p16 and an analysis of a..., Alexander [/bib_ref] [bib_ref] Prevalence of six types of human papillomavirus in inverted papilloma and papillary..., Chan [/bib_ref] [bib_ref] High-risk human papillomavirus DNA detected in primary squamous cell carcinoma of urinary..., Chapman-Fredricks [/bib_ref] [bib_ref] Prevalence of high-risk human papillomavirus in primary squamous cell carcinoma of urinary..., Collins [/bib_ref] [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Detection of human papillomavirus DNA sequences in cancer of the urinary bladder..., Gopalkrishna [/bib_ref] [bib_ref] Human papillomavirus and p16 expression in inverted papillomas of the urinary bladder, Gould [/bib_ref] [bib_ref] Human papilloma virus and p53 expression in bladder cancer in Egypt: Relationship..., Tel [/bib_ref] [bib_ref] Human papillomavirus DNA and abnormal p53 expression in carcinoma of the urinary..., Kamel [/bib_ref] [bib_ref] Detection of human papillomavirus infection and p16 immunohistochemistry expression in bladder cancer..., Kim [/bib_ref] [bib_ref] Human papillomavirus infection and transitional cell carcinoma of the bladder. Immunohistochemistry and..., López-Beltrán [/bib_ref] [bib_ref] Association of human papilloma virus (HPV) infection with oncological outcomes in urothelial..., Moghadam Ohadian [/bib_ref] [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] [bib_ref] Condyloma Acuminatum of Urinary Bladder: Relation to Squamous Cell Carcinoma, Samarska [/bib_ref] [bib_ref] Is there a correlation between HPV and urinary bladder carcinoma?, Shaker [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Etiological correlation of human papillomavirus infection in the development of female bladder..., Shigehara [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] [bib_ref] Human papillomaviruses 6/11, 16/18 and 31/33/51 are not associated with squamous cell..., Westenend [/bib_ref] [bib_ref] Does human papillomavirus play a role in the development of bladder transitional..., Youshya [/bib_ref] [fig_ref] Table 4: Overall HPV prevalence among the selected studies [/fig_ref]
## Outcomes
Pooled HPV prevalence was 19% (95% CI: 13%-26%; I2: 96.4%) [fig_ref] Figure 2: Forest plot of HPV pooled prevalence in bladder cancer [/fig_ref] ranging from 0% [bib_ref] Human papillomavirus is not an etiologic agent of urothelial inverted papillomas, Alexander [/bib_ref] [bib_ref] Investigation of human papillomavirus in bladder cancer in a series of Tunisian..., Ben Selma [/bib_ref] [bib_ref] Human papillomavirus and schistosomiasis associated bladder cancer, Cooper [/bib_ref] [bib_ref] Human papillomavirus infection and transitional cell carcinoma of the bladder. Immunohistochemistry and..., López-Beltrán [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Low frequency of human papillomavirus infection in initial papillary bladder tumors, Simoneau [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref] [bib_ref] Role of human papillomavirus in the development of urothelial carcinoma, Yavuzer [/bib_ref] to 83% [bib_ref] Human papilloma virus is not detectable in samples of urothelial bladder cancer..., Schmid [/bib_ref]. 619/3682 (16.8%) BCa samples were positive [fig_ref] Table 4: Overall HPV prevalence among the selected studies [/fig_ref].
No risk differences were found between females and males [pooled risk difference (95% CI): 0.0046 (−0.0545; 0.0636); p-value: 0.87999; I2: 13.5%] [fig_ref] Figure 1: PRISMA 2020 flow diagram for new systematic reviews, which included searches of... [/fig_ref]. No statistically significant risk differences were found when prevalences related to patients in stage ≤T1 and ≥T2 stage were compared [pooled risk difference (95% CI) = −0.0659
# Discussion
This systematic review was performed to evaluate the prevalence of HPV infection in BCa, keeping into consideration confounding demographic, histological, and diagnostic variables. [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Human papillomavirus and p16 expression in inverted papillomas of the urinary bladder, Gould [/bib_ref] [bib_ref] Detection of human papillomavirus infection and p16 immunohistochemistry expression in bladder cancer..., Kim [/bib_ref] [bib_ref] Human Papillomavirus in Urothelial Carcinoma of Bladder: An Indian study, Mete [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Is there a correlation between HPV and urinary bladder carcinoma?, Shaker [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Overexpression of p16(INK4a) in urothelial carcinoma in situ is a marker for..., Steinestel [/bib_ref] [bib_ref] Human papillomavirus associated with bladder carcinoma? Analysis by polymerase chain reaction, Tekin [/bib_ref]. Also, 69/479 (6.9%) were multiple infections [bib_ref] Koilocytosis: Correlations with high-risk HPV and its comparison on tissue sections and..., Aggarwal [/bib_ref] [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Human papillomavirus and p16 expression in inverted papillomas of the urinary bladder, Gould [/bib_ref] [bib_ref] Human papillomavirus DNA and abnormal p53 expression in carcinoma of the urinary..., Kamel [/bib_ref] [bib_ref] Analysis of p53 tumor suppressor gene mutations and human papillomavirus infection in..., Kim [/bib_ref] [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Low frequency of human papillomavirus infection in initial papillary bladder tumors, Simoneau [/bib_ref] [bib_ref] p53 overexpression and human papillomavirus infection in transitional cell carcinoma of the..., Tenti [/bib_ref] [bib_ref] Human Papillomavirus Prevalence and Integration Status in Tissue Samples of Bladder Cancer..., Yan [/bib_ref]. The most prevalent genotype was HPV-16 (216/479; 45.1%), followed by HPV-18 (153/479; 31.9%), HPV-6 (25/479; 5.2%), and HPV-11 (17/479; 3.5%). Moreover, 68 (14.2%) infections were caused by other HR-HPV genotypes [fig_ref] Table 5: HPV prevalence stratified by the presence [/fig_ref].
## Case-control studies
The studies with a case-control design numbered 11/46 (23.9%) in the selection, with a total of 611 cases vs. 227 controls [bib_ref] Association Between Human Papillomavirus and Transitional Cell Carcinoma of the Bladder, Abdollahzadeh [/bib_ref] [bib_ref] Role of human papillomavirus types 16, 18, and 52 in recurrent cystitis..., Badawi [/bib_ref] [bib_ref] Correlation between human papillomavirus infection and bladder transitional cell carcinoma, Barghi [/bib_ref] [bib_ref] Human papillomavirus and p16 expression in inverted papillomas of the urinary bladder, Gould [/bib_ref] [bib_ref] Detection of human papillomavirus infection and p16 immunohistochemistry expression in bladder cancer..., Kim [/bib_ref] [bib_ref] Human Papillomavirus in Urothelial Carcinoma of Bladder: An Indian study, Mete [/bib_ref] [bib_ref] Is There any Association between Urothelial Carcinoma of the Bladder and Human..., Sarier [/bib_ref] [bib_ref] Is there a correlation between HPV and urinary bladder carcinoma?, Shaker [/bib_ref] [bib_ref] Etiologic role of human papillomavirus infection in bladder carcinoma, Shigehara [/bib_ref] [bib_ref] Overexpression of p16(INK4a) in urothelial carcinoma in situ is a marker for..., Steinestel [/bib_ref] [bib_ref] Human papillomavirus associated with bladder carcinoma? Analysis by polymerase chain reaction, Tekin [/bib_ref]. The overall prevalence was 27.3% (167/611) and 4.4% (10/227) in cases and controls, respectively. The pooled odds ratio (OR) was 7.8406 (95% CI, 4.3425-14.1567; p-value < 0.00001; I2: 34.7%) for the association between HPV infection and occurrence of BCa [fig_ref] Table 5: HPV prevalence stratified by the presence [/fig_ref] ; [fig_ref] Figure 3: Forest plot of overall odds ratio in case-control studies [/fig_ref].
# Discussion
This systematic review was performed to evaluate the prevalence of HPV infection in BCa, keeping into consideration confounding demographic, histological, and diagnostic variables.
An overall HPV prevalence of 19% was found in 46 studies, in line with previous meta-analyses which reported a prevalence of 16.88% [bib_ref] Human papillomavirus infection and bladder cancer risk: A meta-analysis, Li [/bib_ref] and 14.3% [bib_ref] Bladder cancer and human papillomavirus association: A systematic review and meta-analysis, Khatami [/bib_ref]. Despite the growing number of reported studies, the role of HPV in cancers other than genital, anal, head, and neck cancers is still debated, due to the heterogeneity in the study design, population enrolled, and HPV detection methods. In fact, several systematic reviews, which evaluated the prevalence of infection of DNA-based vs. non-DNA-based methods confirmed the higher specificity and sensitivity of molecular-based methods [bib_ref] Meta-analysis of studies analyzing the relationship between bladder cancer and infection by..., Gutiérrez [/bib_ref] [bib_ref] Meta-analysis of studies analyzing the role of human papillomavirus in the development..., Jimenez-Pacheco [/bib_ref]. Moreover, the use of genotype primers designed for shorter DNA sequences reduced the risk of "false negative" results in comparison with broad-spectrum primers (i.e., GP5+/6+), especially for FFPE specimens often undergoing DNA damage [bib_ref] Human papillomavirus infection and bladder cancer risk: A meta-analysis, Li [/bib_ref]. Therefore, the implementation of a standardized procedure for HPV detection could better clarify the impact of HPV in BCa pathogenesis.
A statistically significant difference in the prevalence of infection was found in different histological subtypes, with higher estimates in SCC than those in TCC and UC. The high affinity of HPV to differentiating squamous epithelium, previously demonstrated in cervical, head and neck, and anogenital carcinomas [bib_ref] The natural history of human papillomavirus infections of the mucosal epithelia, Chow [/bib_ref] , as well as the ability of the virus to evade and inactivate the immune response, could explain the mechanism of carcinogenesis in the bladder epithelium. Since SCC has poor prognosis and is associated with worse outcomes in different sites [bib_ref] Human papillomavirus and urinary bladder cancer revisited, Jørgensen [/bib_ref] , the confirmation of the role of HPV in SCC could improve the epidemiological burden and the prognosis of potential cases.
The assessment of the role of HPV as causative agent in different histological subtypes could have been affected by the adoption of the new WHO BCa classification [bib_ref] WHO Classification of Tumours of the Urinary System and Male Genital Organs-Part..., Humphrey [/bib_ref]. This re-cent recommendation changed the traditional nomenclature of "transitional cell carcinoma" in "urothelial carcinoma", causing potential misinterpretations of the results. On this basis, the identification of standard accurate procedures for the detection and diagnosis of HPV infection could be helpful to estimate the burden of cancers associated with HPV. However, the available molecular techniques show a high diagnostic accuracy.
Although several studies suggested a relationship between advanced stage (i.e., ≥T2) and HPV prevalence [bib_ref] Detection of human papillomavirus DNA in urinary bladder carcinoma by in situ..., De Gaetani [/bib_ref] [bib_ref] Association of human papilloma virus (HPV) infection with oncological outcomes in urothelial..., Moghadam Ohadian [/bib_ref] [bib_ref] Detection of HPV infection in urothelial carcinoma using RNAscope: Clinicopathological characterization, Musangile [/bib_ref] , the poor information collected in the selected studies did not prove this association.
Similarly to other HPV-related cancers, HPV-16 was the most prevalent genotype in BCa, supporting previous findings on the increased risk of BCa in cases of infection caused by high-risk HPV genotypes, mainly by HPV-16 [bib_ref] Human papillomavirus infection and bladder cancer risk: A meta-analysis, Li [/bib_ref]. The high detection rate of HPV-16, together with HPV-18 and the low-risk HPV-6, strongly supports the administration of HPV vaccines to prevent HPV-related cancers in both sexes.
Finally, our study showed a significant association between HPV and BCa (OR: 7.84), confirming the findings of recent meta-analyses on the risk of BCa [bib_ref] Human papillomavirus infection and bladder cancer risk: A meta-analysis, Li [/bib_ref] [bib_ref] Meta-analysis of studies analyzing the relationship between bladder cancer and infection by..., Gutiérrez [/bib_ref] [bib_ref] Human papillomavirus and urinary bladder cancer revisited, Jørgensen [/bib_ref]. However, these results are in contrast with a previous study published by al. [bib_ref] Bladder cancer and human papillomavirus association: A systematic review and meta-analysis, Khatami [/bib_ref] , who described a non-significant association between infection and cancer and highlighted the scientific need of larger case-control studies. Although the majority of case-control studies were classified as "medium-quality", controls showed a prevalence of infection <10%, regardless of HPV detection methods, suggesting the etiological role of HPV in BCa.
Previous systematic reviews investigated the role of HPV infection in BCa and described a moderate association. Our study selection and analysis showed an important role of HPV infection in SCC BCa. However, some limitations should be acknowledged. More stratified analyses related to demographic (i.e., geographical area, behavioral factors, occupational exposures) and clinical (i.e., stage, HPV detection methods, prognosis of patients) confounders should be performed. The cross-sectional assessment does not help prove the temporal relationship between an exposure and the development of cancer.
# Conclusions
The present systematic review highlights a potential role of HPV in the development of bladder cancer, indirectly supporting the adoption of primary preventive strategies in both sexes, as recommended by international authorities. Further epidemiological studies are needed to confirm those findings and assess the role of diagnostic and preventive strategies for HPV-related bladder cancer.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/diagnostics12071759/s1, Figures S1-S3, Tables S1-S5. Institutional Review Board Statement: Ethical review and approval were waived due to the design of the study.
Informed Consent Statement: Patient consent was waived due to the design of the study.
# Data availability statement:
The data is available in case it is requested for motivated reasons.
## Conflicts of interest:
The authors declare no conflict of interest.
[fig] Figure 1: PRISMA 2020 flow diagram for new systematic reviews, which included searches of databases and registers only. [/fig]
[fig] 19, Figure 2: −0.173; 0.0411); p-value: 0.22746; I2: 45.2%], as well as in patients with G1 and G2/G3 tumors, [pooled risk difference (95% CI): −0.0451 (−0.1447; 0.0546); p-value: 0.37542; I2: 75.9%] (Figures S2 and S3). Pooled prevalence stratified by histological subtypes was 36.5% (95% CI: 15.9-60.1%; I2: 88.2%), 32.5% (95% CI: 23.8-41.8%; I2: 90.7%), and 18.5% (95% CI: 3.8-40.9%; I2: 97.3%) for SCC, TCC, and UC, respectively(Figure 3), with a statistically significant difference between the prevalence in SCC vs. TCC (p-value: 0.0002) and vs. UC (p-value < 0.0001).Diagnostics 2022, 12, x FOR PEER REVIEW 12 of Forest plot of HPV pooled prevalence in bladder cancer. Adapted from. [/fig]
[fig] Figure 2: Forest plot of HPV pooled prevalence in bladder cancer. Adapted from. [/fig]
[fig] Figure 3: Forest plot of overall odds ratio in case-control studies. Adapted from [/fig]
[fig] Author: Contributions: Conceptualization, N.M., B.D.L., L.S. and G.S.; methodology, N.M., B.D.L., L.S., A.P. and G.S.; software, B.D.L., L.S. and G.S.; validation, N.M., B.D.L., L.S., I.S., A.P. and G.S.; formal analysis, B.D.L., L.S. and G.S.; investigation, N.M., B.D.L., G.S.; resources, N.M., B.D.L., L.S., I.S., A.D.R., A.P. and G.S.; data curation, N.M., B.D.L., L.S., A.P. and G.S.; writing-original draft preparation, N.M., B.D.L., I.S. and A.D.R.; writing-review and editing, N.M., B.D.L., A.P. and G.S.; visualization, N.M., B.D.L., L.S., I.S., A.D.R., A.P. and G.S.; supervision, A.P. and G.S. All authors have read and agreed to the published version of the manuscript. [/fig]
[fig] Funding: Article Processing Charge was funded by Cariplo Foundation-Biomedical Research conducted by Young Researchers 2019. [/fig]
[table] Table 1: Characteristics of the included studies (n = 46).Low prevalence of HPV detection and genotyping innon-muscle invasive bladder cancer using single-step PCR followed by reverse line blotTable 1. Cont. Table 1. Cont. [/table]
[table] Table 3: NOS quality assessment table. Each study was awarded one star per item within the selection and exposure categories. A maximum of two stars could be awarded for comparability. The score is the sum of the awarded stars and ranges from zero to nine. [/table]
[table] Table 4: Overall HPV prevalence among the selected studies. If different prevalence values were outlined in the studies, the value from the standard detection technique (or the highest in case of comparison between standard methods) was reported in the table below. [/table]
[table] Table 5: HPV prevalence stratified by the presence (case) or absence (control) of any type of bladder tumor (case-control studies). [/table]
|
XIAP-associating factor 1, a transcriptional target of BRD7, contributes to endothelial cell senescence
X-linked inhibitor of apoptosis (XIAP)-associated factor 1 (XAF1) is well known as an antagonist of XIAP-mediated caspase inhibition. Although XAF1 serves as a tumor-suppressor gene, the role of XAF1 in cellular senescence remains unclear. We found that XAF1 expression was increased by genotoxic agents, such as doxorubicin and ionizing radiation in pulmonary microvascular endothelial cells, consequently leading to premature senescence. Conversely, downregulation of XAF1 in premature senescent cells partially overcame endothelial cell senescence. p53 knockdown, but not p16 knockdown, abolished senescence phenotypes caused by XAF1 induction.XAF1 expression was transcriptionally regulated by Bromodomain 7 (BRD7). XAF1 induction with interferon-gamma (IFN-γ) treatment was abrogated by BRD7 knockdown, which resulted in blocking interferon-induced senescence. In lung cancer cells, XAF1 tumor suppressor activity was decreased by BRD7 knockdown, and inhibition of tumorgrowth by IFN-γ did not appear in BRD7-depleted xenograft tumors. These data suggest that XAF1 is involved in BRD7-associated senescence and plays an important role in the regulation of endothelial senescence through a p53-dependent pathway. Furthermore, regulation of the BRD7/XAF1 system might contribute to tissue or organismal aging and protection against cellular transformation.
# Introduction
Normal somatic cells cultured in vitro have a limited ability to divide before entering a state of irreversible proliferative arrest termed replicative senescence [bib_ref] The serial cultivation of human diploid cell strains, Hayflick [/bib_ref]. Replicative senescence is induced by telomere attrition, response to activated oncogenes and DNA damage, deregulated nutrient sensing, loss of proteostasis and epigenetic alterations during aging [bib_ref] The hallmarks of aging, Lopez-Otin [/bib_ref] [bib_ref] The asymmetry of telomere replication contributes to replicative senescence heterogeneity, Bourgeron [/bib_ref] [bib_ref] Replicative senescence is associated with nuclear reorganization and with DNA methylation at..., Hanzelmann [/bib_ref].
Irreversible growth arrest is induced in primary cells through the expression of activated oncogenes such as Ras [bib_ref] Oncogenic ras provokes premature cell senescence associated with accumulation of p53 and..., Serrano [/bib_ref] or by activation of tumor suppressor genes [bib_ref] Significant role for p16INK4a in p53-independent telomere-directed senescence, Jacobs [/bib_ref]. Numerous studies have implicated the tumor suppressors p53, p16 and Rb as common major effectors of cellular senescence in normal somatic cells [bib_ref] Reversal of human cellular senescence: roles of the p53 and p16 pathways, Beausejour [/bib_ref] [bib_ref] Senescent cells, tumor suppression, and organismal aging: good citizens, bad neighbors, Campisi [/bib_ref]. In this study, we used genetic approaches to search for previously unexplored senescence regulators in endothelial cells, particularly those involved in BRD7-associated senescence.
Bromodomain 7 (BRD7) is a unique component of the SWI/SNF polybromo-associated BRG1-associated factor (PBAF) complex that contributes to proliferation regulation [bib_ref] BRD7 suppresses the growth of Nasopharyngeal Carcinoma cells (HNE1) through negatively regulating..., Peng [/bib_ref]. It was originally identified as a gene whose mRNA was downregulated in nasopharyngeal carcinoma [bib_ref] BRD7, a novel PBAF-specific SWI/SNF subunit, is required for target gene activation..., Kaeser [/bib_ref]. Recent studies have implicated BRD7 as a regulator of replicative senescence based on the induction of resistance to Ras-induced senescence by BRD7 depletion [bib_ref] BRD7 is a candidate tumour suppressor gene required for p53 function, Drost [/bib_ref] or the induction of oncogene-induced senescence through BRD7 interaction with p53 and p300 [bib_ref] Polybromoassociated BRG1-associated factor components BRD7 and BAF180 are critical regulators of p53..., Burrows [/bib_ref]. Bromodomains are evolutionally conserved domains that have specific binding affinity for acetylated lysines on histone N-terminal tails [bib_ref] Acetyllysinebinding and function of bromodomain-containing proteins in chromatin, Dyson [/bib_ref]. Although the function of bromodomains still requires further investigation, bromodomain proteins modulate chromatin remodeling and modification, thereby facilitating accession of transcription factors to chromatin [bib_ref] The structural basis for the recognition of acetylated histone H4 by the..., Owen [/bib_ref] [bib_ref] The bromodomain: a chromatin browser?, Filetici [/bib_ref] [bib_ref] The bromodomain: a regulator of ATP-dependent chromatin remodeling?, Horn [/bib_ref].
X-linked inhibitor of apoptosis (XIAP)-associated factor 1 (XAF1) directly and indirectly regulates p53mediated apoptosis as a tumor suppressor gene. XAF1 is expressed ubiquitously in all healthy adult and fetal Oncotarget 5119 www.impactjournals.com/oncotarget tissues, but is lost or reduced in a variety of cancer cell lines because of the aberrant promoter hypermethylation of its gene [bib_ref] Expression and genetic analysis of XIAP-associated factor 1 (XAF1) in cancer cell..., Fong [/bib_ref] [bib_ref] Identification of XAF1 as an antagonist of XIAP anti-Caspase activity, Liston [/bib_ref]. XAF1 was originally identified as a nuclear protein that has the ability to bind XIAP and antagonize the ability of XIAP to suppress caspase activity and cell death [bib_ref] Identification of XAF1 as an antagonist of XIAP anti-Caspase activity, Liston [/bib_ref]. XAF1 can also induce apoptosis through an alternative pathway by enhancing TNF-alpha independently of interaction with XIAP [bib_ref] Xaf1 can cooperate with TNFalpha in the induction of apoptosis, independently of..., Xia [/bib_ref].
Despite previous reports showing the implications of XAF1 in p53-mediated apoptosis in cancer, the molecular and cellular effects of XAF1 in primary normal vascular endothelial cells have not been examined. In the current study of the transcriptional regulation by BRD7 in endothelial cell senescence during irradiation, we have found a correlation between XAF1 and BRD7 in radiationinduced senescence. In this study, we demonstrate that XAF1 plays a crucial role in cellular senescence through transcriptional regulation by BRD7 in human endothelial cells.
# Results
## Xaf1 expression increases during dna damageinduced senescence in endothelial cells
To investigate whether XAF1 is associated with cellular senescence in pulmonary endothelial cells, we examined XAF1 expression levels in young and old cells by semi-quantitative PCR (q-PCR), real-time PCR and Western blot analysis. Senescent cells are known to be resistant to mitogen-induced proliferation, express SAβ-gal and have a characteristically enlarged and flattened morphology. Using serial passaging with trypsinization, senescent cells (also referred to herein as 'old cells') were obtained and characterized by p53/p21 activation and SAβ-gal staining , 1B). XAF1 protein levels were upregulated 3-fold or more in the old endothelial cells : Increase in XAF1 expression in cellular senescence. A. XAF1 protein expression levels XAF1 were analyzed by Western blotting with antibodies against XAF1 and GAPDH (loading control) in aged HMVECs. The levels of cyclin A, phospho-p53(S15), p53, p21 and β-actin proteins were detected by Western blot analysis. The figure shows representative data from three independent experiments. b. Percentages of SA-β-gal positive cells analyzed in young and old HMVECs. **p < 0.01 versus the young cells. c. XAF1 and p53 protein expression levels were detected by Western blot analysis. d. HMVECs treated with doxorubicin (Doxo, 1 µM) or ionizing radiation (IR, 4 Gy). The cells were incubated for 0, 24 or 48 h, and the XAF1 mRNA expression levels were measured by semi-quantitative PCR and real-time PCR analyses. Values are expressed as the mean ± SD of three independent experiments. **p < 0.01 versus the control group without Doxo or IR treatment. e. After treatment with Doxo or IR for 4 days, Percentages of SA-β-gal positive cells analyzed in young and premature senescent cells. *p < 0.05 versus the control cells. Y, young cells; O, old cells.
Oncotarget 5120 www.impactjournals.com/oncotarget . XAF1 expression was increased by DNA damaging agents, such as doxorubicin (Doxo) and ionizing radiation (IR) [bib_ref] DNA-SCARS: distinct nuclear structures that sustain damage-induced senescence growth arrest and inflammatory..., Rodier [/bib_ref] [bib_ref] Role of epigenetics in human aging and longevity: genome-wide DNA methylation profile..., Gentilini [/bib_ref] , which induced cellular senescence in endothelial cells . These cells showed senescent phenotypes that distinguished them from the early-passage cells. Doxo or IR treatment increased XAF1 protein and mRNA levels in a time-dependent manner , 1D) and resulted in morphological changes in old cells and increases in SA-β-gal staining activity . These results imply that XAF1 upregulation is implicated in cellular senescence in old and IR-treated endothelial cells.
## Effects of xaf1 upregulation on cellular senescence in hmvecs
Because XAF1 expression levels were increased in senescent cells and by treatment with the DNA damagerelated senescence inducers Doxo and IR, we tested whether XAF1 overexpression has an impact on cellular senescence in HMVECs. HMVECs were transduced with XAF1 lentivirus, and senescence markers in XAF1overexpressing cells were examined. XAF1 upregulation caused a decrease in cell proliferation and an increase in SA-β-gal staining compared with the control lentivirustransduced cells [fig_ref] Figure 2: effects of XAF1 on the upregulation of cellular senescence in young HMVecs [/fig_ref] , 2C). Increased XAF1 expression induced the expression of cell cycle regulators, such as p53 and p21, and inhibited the expression of Cyclin A [fig_ref] Figure 2: effects of XAF1 on the upregulation of cellular senescence in young HMVecs [/fig_ref] , resulting in G0/G1 cell-cycle arrest but not apoptosis [fig_ref] Figure 2: effects of XAF1 on the upregulation of cellular senescence in young HMVecs [/fig_ref]. Taken together, these results suggest that XAF1 might play an important role in inducing senescence in HMVECs.
## Xaf1 knockdown reverses premature senescence in hmvecs
To investigate the role of XAF1 in cellular senescence, HMVECs were treated with two DNA damage with XAF1 or negative control lentiviruses and incubated for 3 days. XAF1 mRNA expression levels were measured by semi-quantitative PCR and XAF1, cyclin A, p53 and p21 protein levels were detected by Western blot analysis. b. Cell proliferation was measured by cell counting and c. the percentages of SA-β-gal positive cells were analyzed. *p < 0.05 and ** < 0.01 versus the vector group. d. XAF1transduced cells were stained with Annexin V-FITC and propidium iodide (PI). To measure apoptosis, fluorescence intensities of Annexin V-FITC were analyzed by flow cytometry. Etoposide (EP, 100 µM) was used as a positive treatment. Cleaved caspase 3 activity was analyzed by Western blot. e. Cell cycle profiles were analyzed by PI staining and flow cytometry. Values are expressed as the mean ± SD of three independent experiments. Representative data from three independent experiments are shown. *p < 0.05 and **p < 0.01 versus the vector group. V, control vector lentivirus-transduced cells; XAF1, XAF1 lentivirus-transduced cells.
Oncotarget 5121 www.impactjournals.com/oncotarget agents (Doxo and IR) that trigger premature senescence. XAF1 mRNA and protein levels were downregulated by gene silencing using XAF1 siRNAs in the senescent cells. The Doxo-and IR-induced increase in p53 expression and activation was reduced in the XAF1 knockdown cells [fig_ref] Figure 3: Partial reverse of premature senescence by XAF1 downregulation in HMVecs [/fig_ref]. Decreased cell proliferation in Doxoor IR-treated cells was partially recovered by XAF1 downregulation [fig_ref] Figure 3: Partial reverse of premature senescence by XAF1 downregulation in HMVecs [/fig_ref]. Repression of XAF1 levels in senescent cells caused morphological changes similar to young cells and a decrease in SA-β-gal activity [fig_ref] Figure 3: Partial reverse of premature senescence by XAF1 downregulation in HMVecs [/fig_ref]. Additionally, XAF1 downregulation in premature senescent cells decreased the population of cells in the G1 phase and increased the population of cells in the S and G2/M phases [fig_ref] Figure 3: Partial reverse of premature senescence by XAF1 downregulation in HMVecs [/fig_ref]. These results suggest that XAF1 knockdown in old cells partially reverses senescence phenotypes.
## Xaf1 induces cellular senescence through a p53 signaling pathway
The p53 and Rb tumor suppressor pathways play critical roles in activating senescence response [bib_ref] The signals and pathways activating cellular senescence, Ben-Porath [/bib_ref]. It is reasonable to expect that XAF1 activates specific signaling pathways to engage p21 via p53 protein and/or p16/Rb proteins. To determine which pathway is involved in XAF1-mediated cellular senescence, we knocked down p16 and p53 using shRNA retroviruses in HMVECs and measured the effects of XAF1 on cellular senescence [fig_ref] Figure 4: effects of p53 or p16 knockdown on cell growth arrest induced by... [/fig_ref]. The p16 knockdown cells showed a decrease in cell proliferation by overexpressing XAF1 similar to the control cells. In contrast, XAF1 overexpression had no effect on cell proliferation in the p53 knockdown cells [fig_ref] Figure 4: effects of p53 or p16 knockdown on cell growth arrest induced by... [/fig_ref]. By measuring the SA-β-gal activity, we demonstrated that p53 knockdown inhibited XAF1induced cellular senescence but p16 knockdown did not [fig_ref] Figure 4: effects of p53 or p16 knockdown on cell growth arrest induced by... [/fig_ref]. Therefore, these results suggest that XAFinduced cellular senescence might be mediated through a p53-dependent pathway.
the role of brd7 in XAF1 transcriptional regulation BRD7 acts a transcriptional cofactor of promoter activity for a subset of p53 target genes [bib_ref] Gene regulation and tumor suppression by the bromodomain-containing protein BRD7, Mantovani [/bib_ref]. We recently identified BRD7 as a potential regulator of DNA damageinduced senescence in endothelial cells. Using DNA microarray analysis, we compared the gene expression profiles from BRD7 down-or upregulated cells to identify Oncotarget 5122 www.impactjournals.com/oncotarget DNA damage-related genes transcriptionally regulated by BRD7. As shown in [fig_ref] Figure 5: transcriptional regulation of XAF1 by brd7 [/fig_ref] , we selected several genes that were potentially regulated by BRD7 and confirmed the expression of each gene in BRD7 down-or upregulated HMVECs using real time PCR. One of these genes, XAF1, was verified to be a BRD7-regulated gene. Interferon regulatory factor 1 (IRF1) is known to be an XAF1 transcription factor and is induced by interferon stimulation. Similar to cells treated with interferon-gamma (IFN-γ), BRD7 could induce XAF1 expression [fig_ref] Figure 5: transcriptional regulation of XAF1 by brd7 [/fig_ref]. To test whether BRD7 exerts a direct influence on XAF1 expression, we measured XAF1 expression when BRD7-downregulated cells were treated with IFN-γ. Despite the presence of IFN-γ stimulation, XAF1 was downregulated by BRD7 depletion, suggesting that BRD7 may mediate a direct effect on XAF1 expression [fig_ref] Figure 5: transcriptional regulation of XAF1 by brd7 [/fig_ref]. Using chromatin immunoprecipitation (ChIP) [fig_ref] Figure 5: transcriptional regulation of XAF1 by brd7 [/fig_ref] , 5E) and luciferase assays [fig_ref] Figure 5: transcriptional regulation of XAF1 by brd7 [/fig_ref] , we further confirmed transcriptional regulation of XAF1 by the BRD7 protein. Consequently, the BRD7 protein increased the promoter activity of the XAF1 gene. Therefore, BRD7 is required for XAF1 transcriptional expression. BRD7 siRNAs in HMVECs. BRD7, CMPK2, ASPM, STIL, SQLE, PLAC8, RIOK2, XAF1, CFB, CBF4 and DTX3L expression levels were analyzed by real time-PCR. Each of the expression levels were normalized to GAPDH expression levels. b. Cells were transduced with BRD7 lentivirus or treated with 0, 20 or 100 ng/ml interferon-gamma (IFN-γ) for 24 h. XAF1, BRD7, IRF1, phospho-p53 and Ac-H3K9 protein levels were detected by Western blot analysis. c. Cells were transfected with BRD7 siRNAs and treated with IFN-γ for 24 h. XAF1, BRD7, IRF1 and phsopho-p53 protein levels were detected by Western blot analysis. d. Cells were transduced with BRD7 lentivirus or transfected with BRD7 siRNAs. Each of cells was treated with 4 Gy IR for 24 h. Chromatin immunoprecipitation (ChIP) assay was performed using the anti-BRD7 antibody and the immunoprecipitated DNA was amplified using primers for p53, IRF1 or XAF1. e. The cells were transfected with the luciferase reporter constructs containing the XAF1 promoter. XAF1 promoter activity in BRD7-expressed cells was determined by luciferase analysis. *p < 0.05 and **p < 0.01 versus the siRNA control group.
Oncotarget 5124 www.impactjournals.com/oncotarget XAF1 as a tumor suppressor gene in lung cancer cells XAF1 is a tumor suppressor that is frequently silenced in many human cancers, but it can inhibit cell proliferation and induce apoptosis in some cancer types. XAF1 increases sensitivity to IR in lung cancer cells, but its effect in lung cancer remains unclear. We tested the cellular effects of XAF1 expression in the human lung cancer cell line A549. Similar to its effect in HMVECs, XAF1 expression increased p21 expression [fig_ref] Figure 6: effects of XAF1 expression in lung cancer cells [/fig_ref] , induced cell cycle arrest [fig_ref] Figure 6: effects of XAF1 expression in lung cancer cells [/fig_ref] and increased SAβ-gal staining [fig_ref] Figure 6: effects of XAF1 expression in lung cancer cells [/fig_ref] in A549 cells. To investigate whether BRD7 acts as critical regulator of XAF1 in lung cancer cells, we examined migration and invasion by transfecting BRD7 siRNAs into cells overexpressing XAF1. XAF1 decreased migration and invasion ability in lung cancer cells, although the increase in XAF1 expression, migration and invasion was not reduced by BRD7 depletion [fig_ref] Figure 6: effects of XAF1 expression in lung cancer cells [/fig_ref]. Unlike in HMVECs, XAF1 induced apoptosis in lung cancer cells [fig_ref] Figure 6: effects of XAF1 expression in lung cancer cells [/fig_ref]. Therefore, these results suggest that BRD7 expression plays a critical role in XAF1 tumor suppression. Oncotarget 5125 www.impactjournals.com/oncotarget effects of brd7-regulated XAF1 in animal model tumors As shown in [fig_ref] Figure 5: transcriptional regulation of XAF1 by brd7 [/fig_ref] , XAF1 is regulated by IFN-γ, meaning that IFN-γ is a good tool for inducing XAF1. Because IFN-γ-induced XAF1 is a potential radiosensitizer in lung cancer cells [bib_ref] Interferon-gamma enhances radiation-induced cell death via downregulation of Chk1, Kim [/bib_ref] , we tested whether BRD7 can affect the radio-sensitizing activity of IFNγ-induced XAF1. After IFN-γ treatment, XAF1 protein levels were regulated by BRD7 depletion . Following IR, a colony forming assay showed that IFN-γ treatment decreased clonogenic cell survival in A549 lung cancer cells but that BRD7 depletion inhibited the decrease in clonogenic cell survival induced by IFN-γ treatment . With INF-γ treatment, we confirmed that BRD7 regulates the tumor suppression activity of XAF1 in vivo. A549 cells were transduced with BRD7 or control shRNA lentiviruses, selected with puromycin and then established as A549 xenografts in : effects of brd7-dependent XAF1 expression in vivo. A. A549 cells were transfected by BRD7 siRNA (siBRD7) or scramble siRNA and then each of the cells was treated with IFN-γ (100ng/ml). Expression of XAF1 and BRD7 proteins were confirmed by Western blotting. b. Each of the cells was irradiated with 0, 1, 2 and 4 Gy of IR, and incubated about 2 weeks for colony formation. Cell survival values were normalized to those of the unirradiated cells. *p < 0.05 versus BRD7-depleted group. c. Effect of IFN-γ administration on growth of A549 tumor xenografts in nude mice. Nude mice (n = 10 per group) were implanted subcutaneously with A549 and groups of mice received vehicle or IFN-γ (1 x 10 5 unit) three times for a week. Tumor size was assessed by direct caliper measurement. Data are means ± SE, n = 10. Tumor volume between shCon and shBRD7 in IFN-γ treated group was statistically significantly different at *p < 0.05 by two-sided Student's t-test. d. Xenograft tumors were embedded with paraffin, were serially sectioned and stained with hematoxylin and eosin (H&E). The tissue samples were immunostained with XAF1 and BRD7 antibodies and observed with fluorescence microscope.
Oncotarget 5126 www.impactjournals.com/oncotarget nude mice. The change in tumor volume was monitored three times a week after INF-γ treatment. Tumor growth was significantly suppressed by INF-γ treatment, but BRD7 depletion prevented the decrease in tumor growth by INF-γ treatment . Xenograft tumors were processed for histology and stained with hematoxylin and eosin (H&E) or immunostained with XAF1 and BRD7 antibodies . H&E staining demonstrated that tumor necrosis induced by IFN-γ treatment was decreased in the BRD7-depleted group and that IFN-γ-induced XAF1 expression was not shown in BRD7-depleted tumors. These observations indicate that INF-γ-induced XAF1 and that XAF1 tumor suppression is determined by BRD7 in vitro as well as in vivo.
# Discussion
Along with surgery and chemotherapy, radiation therapy (radiotherapy) is one of the most important methods for cancer treatment. Currently, more than half of all cancer patients receive radiation therapy. Radiation kills cancer cells by damaging their DNA, but it can also damage normal cells and some tissue types. It is not easy to balance the two conflicting effects during cancer therapy. Exposure to ionizing radiation leads to adverse physiological responses, such as cancer, gastrointestinal failure, bone marrow failure and accelerated aging, but these risks are not exclusively a result of radiotherapy. Generally, the radiation-associated excess relative risks are lower than excess absolute risks in medically exposed individuals [bib_ref] Risks associated with ionizing radiation, Little [/bib_ref]. Therefore, it is important to improve radiotherapy to increase the curative and survival rates for patients.
Thus, we have to consider endothelial cell function in the opposite point of view in cancer and vascular diseases. In cancer biology, endothelial cells are homeostatic factors for regulating angiogenesis-dependent tumor growth [bib_ref] Tumor response to radiotherapy regulated by endothelial cell apoptosis, Garcia-Barros [/bib_ref]. Microvascular damage regulates tumor cell response to radiation, and radiation-induced vascular damages enhance rapid apoptosis in tumors and prevent the recurrence of tumors after radiation treatment [bib_ref] Radiation-induced vascular damage in tumors: implications of vascular damage in ablative hypofractionated..., Park [/bib_ref]. In this study, increased XAF1 was associated with the induction of a variety of senescent phenotypes, including growth arrest by p53 activation and an increase in senescence-associated β-galactosidase (SA-β-gal) staining in both pulmonary endothelial cells and lung cancer cells [fig_ref] Figure 2: effects of XAF1 on the upregulation of cellular senescence in young HMVecs [/fig_ref]. Therefore, XAF1 may be a potential inhibitor of angiogenesis and tumor growth because XAF1 binds directly to p53 and induces p53-mediated apoptosis in cancer.
In contrast, senescent cells have been suggested to contribute to aging-associated pathologies via the accumulation of non-dividing cells and alterations in tissue structure and function [bib_ref] Senescent cells, tumor suppression, and organismal aging: good citizens, bad neighbors, Campisi [/bib_ref]. Therefore, endothelial cell senescence contributes to the age-related decline of vessel structure and the progression of various age-associated vascular diseases. XAF1 induces endothelial cell senescence, suggesting that XAF1 may contribute to vascular aging and the development of aging-associated vascular diseases.
Recently, several groups have reported that radiation induces senescence in endothelial cells. Endothelial cell senescence is associated with vascular damage and vascular disease in vivo. DNA damaging agents, such as ionizing radiation, can accelerate the onset of endothelial cell senescence through numerous changes in protein expression [bib_ref] Quantitative proteomic analysis reveals induction of premature senescence in human umbilical vein..., Yentrapalli [/bib_ref]. We have examined the biological networks that are affected by ionizing radiation from epigenetic alteration. Aging is associated with profound epigenetic changes, resulting in extensive remodeling of gene-expression profiles and disturbances in broad epigenomic landscapes. The relationship between epigenetic changes and aging has been confirmed by changes in DNA methylation levels [bib_ref] Impact of aging on DNA methylation, Richardson [/bib_ref] , histone levels and their post-translational modifications [bib_ref] Chromatin structure as a mediator of aging, Feser [/bib_ref] and transcriptional gene silencing by non-coding RNAs [bib_ref] Novel modulators of senescence, aging, and longevity: Small non-coding RNAs enter the..., Grillari [/bib_ref]. Among them, histone modifications directly contribute to nuclear compartmentalization for the formation of chromatin domains, such as heterochromatin and euchromatin, which are transcriptionally silent and active, respectively [bib_ref] Epigenetics and aging, D'aquila [/bib_ref]. Specifically, with respect to histone modifications and aging, increased H3K9 acetylation leads to shortened life span and premature aging-like phenotypes in response to DNA damage [bib_ref] SIRT6 stabilizes DNA-dependent protein kinase at chromatin for DNA double-strand break repair, Mccord [/bib_ref]. Conversely, H3K9 di-and tri-methylation are reduced by DNA damage and replicative aging [bib_ref] Reduced histone biosynthesis and chromatin changes arising from a damage signal at..., O'sullivan [/bib_ref]. Furthermore, H3K4 demethylation can induce the transcriptional silencing of the retinoblastoma target genes in senescent cells [bib_ref] Rb-mediated heterochromatin formation and silencing of E2F target genes during cellular senescence, Narita [/bib_ref] and methylated H3K36 can regulate cell proliferation and senescence via the p53 and Rb pathways [bib_ref] The H3K36 demethylase Jhdm1b/Kdm2b regulates cell proliferation and senescence through p15(Ink4b), He [/bib_ref].
BRD7 is a selected marker for both cellular senescence and DNA damage response for epigenetic studies with radiation-induced senescence. BRD7 participates in chromatin remodeling by binding to acetylated histones and regulates the expression of genes as a transcription factor. Using genechip arrays and chromatin immunoprecipitation analysis, we have newly founded that XAF1 is transcriptionally increased by BRD7 during radiation exposure [fig_ref] Figure 4: effects of p53 or p16 knockdown on cell growth arrest induced by... [/fig_ref].
Although there is not known for epigenetic alteration by BRD7, the reversibility of epigenetic modifications and gene expression by BRD7 inhibition will provide a new possibility for therapeutic interventions in age-related diseases as well as radiation damage regulation.
## Cell culture
HMVECs in EBM-2 supplemented with EGM-2 and A549s in RPMI 1640 were plated at a density of 2 x 10 5 cells per 100 mm culture plate and cultured at 37°C in a 5% CO 2 humidified incubator. When the subcultures reached 80% confluence, serial passaging was performed by trypsinization. For experiments, cells were used in either passage 5-7 (PD < 24; young) or passage 12-15 (PD > 48; old).
## Induction of cellular senescence by adriamycin treatment and γ-ray irradiation
HMVECs and A549s (2 x 10 5 cells) were seeded in 60-mm culture dishes and incubated for 24 h in culture media. The cells were washed with media and treated with 1 µM doxolubicin (Doxo) for 4 h. After discarding the media containing the Doxo, the cells were washed 3 times with media and incubated in culture media for the indicated times. For ionizing radiation (IR) treatment, cells were exposed to γ-rays with a 137 Cs γ-ray source (Atomic Energy of Canada, Ltd.) at a dosage of 3.81 Gy/ min.
## Virus preparation and transduction
For XAF1 overexpression, the XAF1 transcript was cloned into a pLenti6/V5-TOPO vector (Life Technologies Corporation, Carlsbad, CA, USA). Cells were transduced with XAF1 lentivirus for 48 h. p16 and p53 shRNA retroviruses were prepared by transfection of the pRetroSuper-p53sh and pRetroSuper-p16sh vectors. After incubation for 3 days, media were collected and centrifuged at 1,650 g for 10 min. The viral solution was filtered with 0.45-μm filter membranes and concentrated with a Vivaspin 20 (Sartorius, Goettingen, Germany).
## Cell counting
Cells were harvested by trypsin-EDTA treatment and stained with 0.1% trypan blue, and the cell number was determined using a hemocytometer.
## Clonogenic cell survival assay
A clonogenic cell survival assay was performed by following a standard protocol.
## Invasion and migration assay
Cell invasive activities were assessed using a 12-well Transwell ® insert from Corning Life Sciences (Lowell, MA, USA). Cells were seeded in the absence of serum at a concentration of 2 × 10 5 cells in the upper chamber. The lower chamber was filled with culture medium supplemented with 10% FBS. After 24 hours, invaded cells that had passed through the Matrigel were fixed with methanol, stained with 0.1% crystal violet, and counted under a light microscope (200× magnification). For the migration assay, cells were seeded at a density of 1 x 10 6 cells in a 12-well dish. Scratch wounds were created with pipette tips through the confluent cells. Migration/ wound-healing was observed after 24 h.
## Immunofluorescence staining
Cells were fixed with 3.7% para-formaldehyde in PBS for 3 min and permeabilized in PBS containing Oncotarget 5128 www.impactjournals.com/oncotarget 0.5% Triton X-100 for 5 min. The cells were incubated with an antibody specific to XAF1 (1:200) for 2 h and then with Alexa Fluor ® 594 donkey anti-Goat IgG (1:300) (Life Technologies, Carlsbad, CA, USA) for 30 min. The nuclei were stained with 0.1 μg/ml of 4′,6-diamidino-2phenylindole (DAPI) for 5 min. Images were obtained using a fluorescence microscope.
## Senescence-associated β-galactosidase (sa-β-gal) staining
Cells were plated at a density of 4 x 10 4 cells in 35 mm culture dishes, washed with PBS and fixed with 3.7% (v/v) paraformaldehyde in PBS for 3 min at room temperature. The presence of SA-β-gal activity was determined by incubating the cells with a solution containing 40 mM citric acid-sodium phosphate (pH 6.0), 150 mM NaCl, 2 mM MgCl 2 , 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide and 1 mg/ ml X-gal for 17 h at 37°C. The percentage of blue cells per 100 cells observed under a light microscope was measured.
## Total rna extraction and reverse transcriptionpolymerase chain reaction
RNAs were prepared from cells using the TRI solution according to the manufacturer's instructions. cDNA was prepared from RNA (2 µg) using 2.5 µM oligo-dT primers, 1 mM dNTPs and moloney murine leukemia virus (MMLV) reverse transcriptase. XAF1, p53 and p16 were amplified from the cDNA with AccuPower ® PCR PreMix (Bioneer Inc., Daejeon, Korea) and gene-specific oligonucleotides. GAPDH primers were used to estimate the amount of RNA in each sample. PCR products were separated on 1.2% agarose gels, and the DNAs were visualized by SYBR Green staining.
## Protein extraction and western blot analysis
Cells were washed with PBS and lysed in 70 μl of RIPA buffer (25 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM sodium vanadate, 5 mM NaF and protease inhibitor). The proteins were quantified with the bicinchoninic acid (BCA) method using bovine serum albumin as a standard. Proteins (30 µg) were separated by 10% SDSpolyacrylamide gel electrophoresis and then transferred to nitrocellulose membranes. The membranes were incubated overnight at 4°C with specific antibodies. After washing 5 times in Tris-buffered saline containing 1% Tween 20 (TTBS), the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies. The proteins were visualized using enhanced chemiluminescence (GE Healthcare Life Sciences, Buckinghamshire, United Kingdom).
## Chromatin immunoprecipitation
Cells (1 x 10 6 ) were plated on 100 mm-culture plates and incubated under various conditions. Cells were treated with 1% formaldehyde for 10 min at room temperature to crosslink protein-DNA complexes. Cells were harvested, washed three times in 1× PBS and lysed in SDS with 50 mM Tris-HCI (pH 8.1) and 1 mM EDTA. Cells were sonicated for 1 hour, and the supernatants were evenly split for incubation with the primary antibodies. Reactions with a final volume of 1 ml in dilution buffer with 2 µg of the appropriate primary antibody were rotated overnight at 4°C. Then, 60 µl of Protein A beads was added to each reaction and rotated for 1 h at 4°C. The beads were washed three times. The eluted supernatants as well as input DNA samples were then uncrosslinked via incubation at 65°C for 4 h. The DNA was then ethanol precipitated and treated with Proteinase-K for 30 min at 37°C. The samples were then extracted with phenol:chloroform, precipitated in ethanol and resuspended in 20 µl of distilled water. Promoter binding activity was measured by real time-PCR analysis with specific promoter primers. The promoter primer sequences are as follows: (1) XAF1 promoter, (forward primer, 5'-cctgagtagctgggattata-3' and reverse primer, 5'-ctgggcagtctgtcaatatc-3') and (2) IRF1 promoter (forward primer, 5'-aatcccgctaagtgtttgga-3' and reverse primer, 5'-agtggaagagggaagaaggc-3').
## Flow cytometry
Cells were harvested, washed twice with PBS and fixed with 70% ethanol at -20°C for 1 h. After washing the cells with PBS containing 2% FBS and 0.01% CaCl 2 , RNase (0.5 μg/ml) was added and incubated with the cells at 37°C for 30 min. Then, propidium iodide (50 μg/ml) was added and incubated at room temperature for 20 min. The intracellular propidium iodide fluorescence intensity of each population of 10,000 cells was measured using a Becton-Dickinson FACSCalibur flow cytometer.
## In vivo xenograft assay
Nude mice (Balb/c, 5-week-old males) were obtained from Orientbio, Inc., (Seoul, South Korea) and maintained in a laminar air-flow cabinet under specific pathogen-free conditions. A549 cells were transduced with control shRNA virus (shCTR) or BRD7 shRNA virus (shBRD7) and a 0.1 ml suspension of each containing 1 x 10 6 cells was subcutaneously injected into the right flank of each mouse. When the tumor size reached 150 mm 3 for both the shCTR and shBRD7 groups, the mice treated with www.impactjournals.com/oncotarget each shRNA were randomly divided into two groups (10 mice/group). The experimental group of mice received an intratumoral injection of vehicle or interferon-γ (IFN-γ, 1 x 10 5 international units [IU]) solution three times a week. Two perpendicular diameters of the tumors were measured three times per week with a caliper square by the same investigator, and the tumor volume was calculated using the following equation: Tumor volume (V) mm 3 = (smaller diameter) 2 x (larger diameter) x (π/6). The experiment was terminated 5 weeks after tumor cell implantation because the tumors in the control mice began to show signs of possible necrosis. The use of these animals and the experimental procedures were approved by the Institutional Animal Care and Use Committee of the Korea Institute of Radiological and Medical Sciences.
# Statistical analysis
All of the data are presented as the mean ± SD. Student's two-tailed t-test was employed for all of the analyses, and a p value less than 0.05 was considered statistically significant.
## Abbreviations
HMVEC, human pulmonary microvascular endothelial cell; XAF1, X-linked inhibitor of apoptosisassociating factor 1; BRD7, bromodomain 7; IFN-γ, Interferon-gamma; IRF-1, Interferon regulatory factor-1; PD, population doublings; SA-β-gal, senescenceassociated β-galactosidase.
[fig] Figure 2: effects of XAF1 on the upregulation of cellular senescence in young HMVecs. A. Young cells were transduced [/fig]
[fig] Figure 3: Partial reverse of premature senescence by XAF1 downregulation in HMVecs. A. HMVECs treated with doxorubicin (Doxo, 1 µM) or ionizing radiation (IR, 4 Gy). The cells were incubated for 48 h before transfection with XAF1 siRNAs (siXAF1) or negative control siRNAs. XAF1 knockdown in Doxo-or IR-treated cells was confirmed by RT-PCR (upper panel) and Western blot analysis (lower panel). b. Effects of XAF1 knockdown on premature senescence by Doxo or IR in XAF1 siRNAs-transfected cells were examined by cell proliferation and c. SA-β-gal staining (100x). d. Cell cycle profiles were analyzed by PI staining and flow cytometry. The figure shows representative data from three independent experiments. Values are expressed as the mean ± SD of three independent experiments. *p < 0.05 and **p < 0.01 versus the siRNA control group. [/fig]
[fig] Figure 4: effects of p53 or p16 knockdown on cell growth arrest induced by XAF1. A. Cells were transduced with p53or p16-shRNA retroviruses and incubated for 2 days. P53 and p16 knockdown was confirmed by RT-PCR analysis. The p53-or p16downregulated cells were transduced with XAF1 or negative control lentiviruses and incubated for 2 days. Cell proliferation was measured by b. cell counting for 2 days and c. SA-β-gal staining (100x). Values are expressed as the mean ± SD of three independent experiments. Representative data from three independent experiments are shown. *p < 0.05 and **p < 0.01 versus XAF1 overexpression. p53sh, p53-shRNA retrovirus transduced cells; p16sh, p16-shRNA retrovirus transduced cells.Oncotarget 5123 www.impactjournals.com/oncotarget [/fig]
[fig] Figure 5: transcriptional regulation of XAF1 by brd7. A. XAF1 was upregulated with BRD7 lentivirus or downregulated with [/fig]
[fig] Figure 6: effects of XAF1 expression in lung cancer cells. A. XAF1 was upregulated in A549 lung cancer cells and its expression levels were confirmed by semi-quantitative RT-PCR and western blot analysis. Cell proliferation was measured by b. cell counting for 2 days and c. SA-β-gal staining (100x) and the percentage of SA-β-gal positive cells were analyzed in XAF1 upregulated cells c.. *p < 0.05 and **p < 0.01 versus the vector group. d. Cell migration and e. invasion assays were performed with the XAF1 upregulated cells. *p < 0.05 and **p < 0.01 versus XAF1 overexpression. F. Cleaved caspase 3 activity was analyzed by Western blot. Percentage of apoptosis cells was analyzed by flow cytometry with Annexin V-FITC. Etoposide (EP) was used as a positive treatment. *p < 0.05 versus vector group. [/fig]
|
A new mutually reinforcing network node and link ranking algorithm
This study proposes a novel Normalized Wide network Ranking algorithm (NWRank) that has the advantage of ranking nodes and links of a network simultaneously. This algorithm combines the mutual reinforcement feature of Hypertext Induced Topic Selection (HITS) and the weight normalization feature of PageRank. Relative weights are assigned to links based on the degree of the adjacent neighbors and the Betweenness Centrality instead of assigning the same weight to every link as assumed in PageRank. Numerical experiment results show that NWRank performs consistently better than HITS, PageRank, eigenvector centrality, and edge betweenness from the perspective of network connectivity and approximate network flow, which is also supported by comparisons with the expensive N-1 benchmark removal criteria based on network efficiency. Furthermore, it can avoid some problems, such as the Tightly Knit Community effect, which exists in HITS. NWRank provides a new inexpensive way to rank nodes and links of a network, which has practical applications, particularly to prioritize resource allocation for upgrade of hierarchical and distributed networks, as well as to support decision making in the design of networks, where node and link importance depend on a balance of local and global integrity.Ranking individual elements within networks, including both nodes and links, allows for the identification of important subsets of elements for resource prioritization, such as finding the most authoritative webpages related to a search topic in the internet, discovering the most influential people in a social network, evaluating the most cited scientific papers in a citation network, or identifying the most vulnerable components in infrastructure systems (e.g., transportation networks, power grids, and water systems). Quantification of the criticality of network components helps decision makers inform their management strategies. For example, infrastructure designers can set mandatory safety targets or reliability levels for components (e.g., highways and bridges) in a transportation network, despite resources being typically limited. Therefore, computationally efficient measures that can identify or screen out the importance of network nodes and links are required, as standard methods are too computationally demanding.The ranking of network nodes or links addresses the question "Can important nodes or links in a network be meaningfully identified while keeping input data and computational resources low?" To answer this question, several closed-form ranking measures have been proposed to capture the particular features of networks, mainly based on their topology 1-5 . For example, one of the simplest node importance measures in a network is the degree of a node k i , denoting the number of links connected to it. Traditionally, the adjacency matrix A is used to describe the connectivity patterns of nodes and links of a network. The adjacency matrix has also been used as input for advanced eigenvalue and eigenvector analyses, which continue to gain adoption in practice 6 . Spectral network component rankings are sophisticated extensions of fundamental ranking methodologies, such as node degree centrality k i7,8. One of the most successful developments in spectral ranking of network components has been the PageRank algorithm 3 , which forms the basis of Google webpage rankings, where websites are nodes and hyperlinks are the connections among them. PageRank considers the hyperlink weight normalization and web surfing principles based on random walk models 9 . Another popular ranking algorithm is the Hypertext Induced Topic Selection (HITS) algorithm developed by Kleinberg 4 . The HITS algorithm defines two types of nodes in a network, hubs and authorities, and computes the ranking score of them in a mutually reinforcing way. However, both PageRank and HITS are limited to ranking nodes of a network and they do not account for the importance of links. Very few efficient algorithms for the ranking of links exist in the literature.Ranking studies for webpage networks and citation networks have mainly focused on the ranking of nodes because link ranking in such networks is not of practical interest. This is in sharp contrast to certain types of networks, such as transportation networks, whose links are at least as important as nodes, as well as supply chain networks, whose links are important to find the movement of a product or service to the end customer. However, relatively limited research has been performed so far to investigate the ranking of the links of a network. Betweenness centrality of an edge has been used as an approximate indicator of the link's critical role in the passage of network flow10,11. Several other studies rank the links of networks based on algorithms whose premise is that the most vital links in a network are those whose one-by-one removal result in the greatest change of a system-level performance metric, such as decrease of the efficiency, connectivity loss, or increase of the travel time 12,13 . The N-1 criterion in power systems is a practical example of this approach 14 . Even though the measure of performance of the network may vary, the basic idea for prioritizing the components of the network based on the removal of its components is essentially the same across available studies15,16. However, these link-based measures usually depend on the calculation of network performance (e.g., connectivity reliability and network flow) after the removal of the links, which is time-consuming and depends on the choice of the performance measures. For example, for the whole transportation network of the United States, there are tens of thousands of nodes. Hence, the traditional removal strategy (e.g., based on N-1 criteria) is too time-consuming for practical applications.Therefore, new computationally efficient tools, particularly based on the spectral analysis of networks for joint node-link ranking are needed. This link-aware ranking strategy must remain computationally feasible even for large systems, which is a persistent challenge in infrastructure engineering today 16 . This paper develops new ranking indices by building upon the promising spectral properties of networks. First, this study explores the formulation of a ranking index called Wide Rank (WRank) that can rank the nodes and links of a network simultaneously. Building upon this WRank formalism, a Normalized WRank algorithm (NWRank) is proposed in this manuscript. This NWRank index combines the idea of mutual reinforcement and weight normalization into a unified framework. Then, the WRank and NWRank approaches are applied to a broad set of synthesized network models and compared with the PageRank, HITS, eigenvector centrality, and edge betweenness algorithms, as well as the benchmark N-1 criterion based on a network efficiency metric, which is widely used in the network science and engineering fields to study the performance of various kinds of networks.
# Results
WRank algorithm as a building block. Let G (N, M) represent a network consisting of a node set N with n nodes and a link set M with m links. In addition, it is assumed that no node is directly connected to itself and that multiple links do not exist between nodes. In the WRank algorithm proposed here, instead of focusing solely on the ranking of nodes as emphasized in previous studies, each node i ∈ N receives a ranking score x i and each link a ∈ M also receives a ranking score y a . The premise is that an important node should be pointed to by many critical links (this defines the L operation below), and a critical link should point to important nodes (this defines the Q operation below). The mutually reinforcing relationship of nodes and links through the noted operators can be represented in general as follows, = ( ), = ( ).
( ) x y y x L Q 1
where vectors x = (x 1 , x 2 , …, x n ) T and y = (y 1 , y 2 , …, y m ) T contain the ranking score of each node and link, respectively. The mutual reinforcement operations L and Q can be expressed through the following matrix representations:
[formula] = , = . [/formula]
( ) x Wy y where W is a n × m matrix and Z is a m × n matrix whose generic entries are given by
[formula] = ( ) w [/formula]
if link a points to node i otherwise
[formula] 1 0 3 ia = ( ) z [/formula]
if node j is one of the end nodes of link b otherwise where i, j, k, … ∈ N are nodes and a, b, c, … ∈ M are links. It should be noted that W is the transposed matrix of Z, and vice versa. Therefore, once one of them is obtained, the other one can be easily obtained.
The final ranking scores of each node and link can be obtained through an iterative updating process indexed by t. Similar to the HITS algorithm [bib_ref] Authoritative sources in a hyperlinked environment, Kleinberg [/bib_ref] , the ranking vectors x and y need to be normalized so that their sum is equal to 1 before moving to the next iteration t + 1. If x(t) and y(t) are used to denote ranking scores at the t th iteration, the iterative processes to reach the final solutions are:
[formula] ( + ) = ( ) = ( ) ( ) x W y W Zx t t t 1 5 ( + ) = ( + ) = ( ) ( ) y Z x Z Wy t t t 1 1 6 [/formula]
The final solutions x, y converge to the principal eigenvectors of WZ and ZW or to the linear combination of the principal eigenvectors if more than one principal eigenvector exists, similar to HITS-inspired algorithms. It should be noted that although WZ is similar to the adjacency matrix, the diagonal entries are different. The diagonal entries of the adjacency matrix are zeros but the diagonal entries of WZ are non-zeros. The diagonal entries of the WZ matrix are the in-degrees of the corresponding nodes, which makes the WRank algorithm suitable for directed networks automatically. Hence, WZ provides node-to-node connectivity as well as in-degree in the diagonal, while ZW provides link-to-link connectivity as well as link multiplicity in the diagonal.
Proposed NWRank algorithm. Although the WRank algorithms can work well for most types of networks, two problems could arise for certain networks with prestigious nodes or where network flow (or its approximation) is considered in the ranking of nodes and links. First, if a node with a high ranking score is pointed to by many links, then all of those links also obtain high ranking scores. However, this is not necessarily appropriate. The score received by each link from a prestigious node should be diluted by being shared with others. To solve similar problems, PageRank weights each of the out-going hyperlinks from a webpage, p i , by 1/k i , where k i is the degree of node p i ; thus, every link has the same total out-going weight equal to 1/k i . Even though the PageRank algorithm is used successfully in Google, it does not work in the WRank algorithms. It turns out that every link ends up with exactly the same ranking score if each link has the same out-going weight 1/k i . In fact, both the PageRank and HITS algorithms treat all links equally when distributing ranking scores. However, in practice some links are more important than others. A second challenge is that none of the closed-form network measures, except for Betweenness Centrality approximately, capture the role of network flow in the ranking of nodes or links. Consider the situation depicted in [fig_ref] Figure 1: A low degree node that is important for network flow or its... [/fig_ref] where node 1 lies on a path between two communities. Although node 1 has a low degree, it may have considerable impact on the flow between other nodes within the network. In addition, links b and c also have significant influence on the network flow. However, both PageRank and HITS algorithms do not consider the network flow in the ranking of nodes and only nodes that are well-connected with others obtain a high ranking score. Therefore, node 1 has a very low ranking value if PageRank or HITS algorithms are used to rank the nodes.
To address the two problems stated above, this study expands the WRank formulation into a weight-normalized algorithm referred to as NWRank. Instead of dividing the ranking score of the nodes evenly among each incident link, each link obtains a weight proportional to its neighboring nodes' degree and the Betweenness Centrality (BC) of the link. The link will obtain more weight if the adjacent neighbor nodes have a high degree, the BC of the link is high, or both. In the new method, the Z matrix is replaced by the following H matrix:
[formula] α α = ∑ + ( − ) ∑ ( ) h k k BC BC otherwise 1 0 7 bj b q a b q a [/formula]
where h bj is the score distribution from node j ∈ N to link b ∈ M, q is the number of links that connect to node j; k b and k a represent the degree of the end nodes opposite to j of links b and a = 1, …, q, respectively. Also, BC b and BC a are the Betweenness Centrality levels of links b and a = 1, …, q, respectively. The variable α is the weight coefficient. In the case study, equal weights (i.e. α = 0.5) are chosen between the degree and BC as a starting point (unbiased point), acknowledging that connectivity and network flow contribute equally; in practice, methods based on utility theory or other user-based elicitation strategies can be adopted to establish weights. For now, sensitivity studies are conducted to analyze the influence of the weighting coefficients on the ranking results (presented in the following section). It should be noted that the W matrix can also be modified to include the node betweenness and node degree similar to the H matrix. However, the final ranking score of nodes and links did not change much based on the case study networks introduced below. Therefore, in this study the W matrix is kept unaltered. In sum, for the NWRank algorithm, the iterative processes to reach the final solutions is:
[formula] ( + ) = ( ) = () ( ) x W y W Hx t t t 1 8 ( + ) = ( + ) = ( ). ( ) y H x H Wy t t t 1 1 9 [/formula]
Note that the NWRank algorithm combines the most important mutual reinforcement feature of HITS and the most important hyperlink weight normalization feature of PageRank as well as the consideration of approximate network flow into a unified framework to provide an efficient joint node-link ranking. The mutual reinforcement here refers to the node-link relationship instead of the node-node relationship in HITS algorithms, which is one of the novelties of the NWRank algorithm. In addition, the weight normalization approach proposed here assigns more weight to important links based on both the topology of the network and an approximate form of the network flow through it, making the method applicable to different network problems where both nodes and links are critical for system performance, and where a trade-off between local and global integrity is at work, owing to connectivity and flow considerations, respectively.
Application to diverse networks. To evaluate the WRank algorithm and, in particular, the NWRank algorithm, they are used along with PageRank, HITS, eigenvector centrality, and edge betweenness to study several diverse idealized networks. In order to further evaluate the NWRank algorithm, all the above algorithms are also compared with traditional N-1 removal-based strategies. Different metrics can be used to quantify the network performance in the removal strategy depending on the types of networks. For example, travel time may be used in transportation networks and cost may be used in supply chain networks. As conceived, NWRank is a general ranking algorithm that can be used to rank the nodes or links of generic networks. Therefore, a general metric (i.e. network efficiency E 18 as shown in Equation 10) is used for assessment and comparisons here, where N is the number of nodes in the network, and d ij is the shortest distance between pairs of nodes in a network. Network Efficiency E 18 has been used to describe the performance of various kinds of networks [bib_ref] A topological analysis of scientific coauthorship networks, Cardillo [/bib_ref] [bib_ref] An improved model for structural vulnerability analysis of power networks, Chen [/bib_ref] [bib_ref] Network and eigenvalue analysis of financial transaction networks, Kyriakopoulos [/bib_ref] [bib_ref] Worldwide marine transportation network: Efficiency and container throughput, Deng [/bib_ref] [bib_ref] Bridge retrofit prioritisation for ageing transportation networks subject to seismic hazards, Rokneddin [/bib_ref] [bib_ref] Safety analysis of process industry system based on complex networks theory, Jiang [/bib_ref] , as well as the performance of general complex networks [bib_ref] A network efficiency measure for congested networks, Nagurney [/bib_ref] [bib_ref] Cascade and breakdown in scale-free networks with community structure, Wu [/bib_ref] [bib_ref] Geographical constraints to range-based attacks on links in complex networks, Gong [/bib_ref] [bib_ref] Tunable path centrality: Quantifying the importance of paths in networks, Pu [/bib_ref]. Therefore, network efficiency is a popular metric to describe the network performance in the network science and engineering applications.
[formula] ∑ = ( − ) ( ) ≠ E N N d 1 1 1 10 i j ij [/formula]
Real world complex networks are usually the result of integrating several idealized yet topologically diverse networks of different sizes and configurations. For example, the transportation network of the United States consists of states' transportation networks. The states' transportation networks are made up by the cities' transportation networks, and so on. The rule also applies to other networks, such as the internet, social networks, telecommunication networks, water systems and power grids. Therefore, investigating the configuration of the building blocks (i.e. typical small networks with some homogeneous features) found in large networks is informative [bib_ref] Less is more in modeling large genetic networks, Bornholdt [/bib_ref]. Yazdani et al.divided a subset of typical small diverse networks into two main categories that capture different network configurations: (1) hierarchical network models, where the networks are organized (and possibly operated) in a hierarchical fashion with some of the nodes easily identifiable as the master or upstream nodes while most other nodes are flow distribution junctions and downstream nodes; and (2) distributed network models, where the networks have a relatively uniform topology and only a few hubs or single failure-points are present. This study uses these small yet topologically diverse ensembles of networks, along with a few other ideal topologies (e.g., grids, and scale-free networks), to test the performance of the WRank and NWRank algorithms.
Computational experiments with hierarchical networks. The configuration of the hierarchical networks is likely seen within many spatially organized technological and urban infrastructure systems composed of transmission (global) and distribution (local) subsystems. Four hierarchical networks are selected as case studies and their idealized topologies are shown in [fig_ref] Figure 2: Topology of select hierarchical networks [/fig_ref]. The basic information about these four networks is listed in . [fig_ref] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks [/fig_ref] list the node and link ranking results of the four hierarchical networks. Nodes and links that have the same score are not separated by horizontal lines. For simple hierarchical networks (e.g., T8 and B8), the node ranking of the NWRank algorithm is consistent with the WRank, HITS, and eigenvector centrality algorithms. Therefore, the weight normalization of the link has minor influence on the ranking of the nodes for simple hierarchical networks. Also, it can be seen that the node ranking for WRank and NWRank relative to the HITS algorithm across the other hierarchical networks (e.g., C13 and HT15) is almost the same due to their common mutual reinforcement feature, although some differences are observed as these networks start benefiting from a joint node and link ranking perspective. For instance, in C13 which has clear clusters, it is more beneficial to maintain the node linking such clusters (node 1), whereas in HT15 with no such clusters, it is more beneficial to have some nodes enabling network subsets to remain connected (nodes 2 and 3)-as captured by NWRank and not by HITS. The NWRank also performs better than the naïve WRank, HITS, and eigenvector centrality algorithms based on the results from an N-1 removal strategy as well as a perspective of network integrity where local and global connectivity trade-offs are at work. For example, [fig_ref] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks [/fig_ref] show that for node ranking, NWRank obtains the same ranking results with the N-1 removal strategy for the four hierarchical networks. This is in contrast with other node rankings. For example, while the node ranking of B8 for the PageRank algorithm is the same as the other three algorithms, for most other networks it produces different node ranking results which do not reflect the role of network flow in the ranking of nodes and links. Specifically, the PageRank algorithm ranks node 1 as the least important node of C13, which is at odds from the perspective of network connectivity and network flow. This is because for the PageRank Bat network (B8) 8 13 32
Three-cluster network (C13) 13 Hierarchical tree network (HT15) 15 [bib_ref] Long-term effect of the n-1 criterion on cascading line outages in an..., Ren [/bib_ref] 30 . Basic information about selected hierarchical networks.
algorithm, the ranking of nodes is highly correlated with the ranking by the in-degree of nodes [bib_ref] PageRank, HITS and a unified framework for link analysis, Ding [/bib_ref] [bib_ref] Ranking vertices or edges of a network by loops: a new approach, Van Kerrebroeck [/bib_ref] [bib_ref] Ranking stability and super-stable nodes in complex networks, Ghoshal [/bib_ref]. Node 1 has relatively small in-degree in C13, and it appears at the bottom of the ranking list. [fig_ref] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks [/fig_ref] [fig_ref] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks [/fig_ref]. Node and link ranking results for the C13 and HT15 hierarchical networks.
Regarding the ranking of links, NWRank is the only algorithm that obtains the same ranking results with the N-1 removal strategy for most of the hierarchical networks (i.e. T8, C13, and HT15), with a difference in B8 where NWRank favors edges to maintain clusters (edges 2-4 and 8-10) to keep most of the network integrated at the expense of leaving some nodes connected with lower-ranked links (depending on edge 1 in this case, which is favored by the N-1 and betweenness strategies). In addition, except for HT15, WRank and NWRank provide the same link ranking results, where the difference stems from the trade-off mechanism of NWRank to handle local and global connectivity perspectives, where topology dominates the local aspect and flow the global one. Also, the NWRank algorithm obtains the same link ranking results with edge betweenness algorithms for C13 and HT15. Across the four networks, it can be seen that the NWRank algorithm aligns either with the WRank algorithm or the edge betweeness algorithm because it combines the features of both of them, while revealing expected edge rankings each time (i.e., favoring integrity within clusters over entire network connectivity when network leafs of weakly connected subsets are present). This is captured in the importance of links 1, 8, and 15 and node 1 of C13, or links 1 and 2 and nodes 2 and 3 of HT15. The superiority of the NWRank algorithm for the ranking of nodes and links is further confirmed below with other complex networks, such as distributed networks, which tend not to have obvious prestigious nodes.
## Computational experiments with distributed networks.
While it is relatively easy to identify the master nodes or links for the hierarchical networks, it is usually more difficult to identify the authoritative components for distributed networks, such as sensor networks and certain types of transportation and commodity distribution networks. Different from the hierarchical networks, distributed operation is usually self-organized, which allows for a greater allocation of redundancy and operation of alternative supply paths. Four distributed networks are chosen in this study and their topologies are shown in [fig_ref] Figure 3: Topology of select distributed networks [/fig_ref]. The basic information of these networks is listed in [fig_ref] Table 4: Basic information of the selected distributed networks [/fig_ref]. To facilitate ranking comparisons across strategies, the ranking agreement between all the ranking algorithms with the N-1 removal strategy is quantified by cosine similarity 37 for distributed networks.
While [fig_ref] Table 5: Node and link ranking results of the Grid25 and DT10 distributed networks [/fig_ref] list the node and link ranking results of the four distributed networks, [fig_ref] Table 7: Comparison of the node ranking between the ranking algorithms and N-1 removal... [/fig_ref] and synthesize comparisons between all the ranking algorithms and the N-1 removal-based approach. It can be seen that for node ranking, NWRank is the only ranking algorithm that obtains high cosine similarities with the traditional N-1 removal strategy for all the four distributed networks. The cosine similarity values are at least 0.856, indicating good agreement. While other ranking algorithms can obtain similar ranking results with the removal strategy for certain distributed networks, they are not consistent across all networks, especially for the UD16 network as explained later. As with the hierarchical networks, for certain distributed networks the node ranking of WRank and HITS algorithms is very similar (e.g. identical for Grid25, and similar for DT10, SF10, and UD16). In addition, the eigenvector centrality and HITS obtain the same node ranking results for the four distributed networks. Also, based on [fig_ref] Table 7: Comparison of the node ranking between the ranking algorithms and N-1 removal... [/fig_ref] , PageRank still does not perform well for distributed networks. Because NWRank combines the mutual reinforcement feature of HITS and WRank as well as the weight normalization feature of PageRank, the node ranking produced by NWRank is in between the rankings produced by WRank, HITS, eigenvector centrality, and PageRank as shown by the SF20 and UD16 networks.
For the link ranking, the NWRank algorithm is in between the WRank and edge betweenness algorithms because it combines the features of both of them. Based on , the NWRank algorithm still performs better than the WRank algorithm and the edge betweenness algorithms for most of the distributed networks, especially for UD16 network. UD16 is a good example that illustrates the potential problems of the naïve WRank and HITS algorithms. It is easy to identify that three communities exist in the UD16 network. For the link ranking of UD16, the top 11 links produced by the WRank algorithm all belong to the same community while the top 11 links produced by the NWRank algorithm distribute among all the three communities, especially the links that connect the communities, which reflects the importance of approximate network flow. The same is true for the ranking of nodes. The WRank and HITS algorithms lead to very similar rankings and all the top ranking nodes belong to the same communities. However, NWRank (and PageRank) finds the top ranking nodes from all of the three communities. This problem for the ranking of nodes in HITS-inspired algorithms, including the naïve WRank approach, was originally called the Tightly Knit Community (TKC) effect by Lempel and Moran [bib_ref] The stochastic approach for link-structure analysis (SALSA) and the TKC effect, Lempel [/bib_ref]. A tightly knit community is a small but highly interconnected set of nodes and the TKC effect occurs when such community scores high in link-analyzing algorithms [bib_ref] The stochastic approach for link-structure analysis (SALSA) and the TKC effect, Lempel [/bib_ref]. This effect hampers the ability of the mutual reinforcement approach to identify meaningful authorities [bib_ref] Ranking stability and super-stable nodes in complex networks, Ghoshal [/bib_ref]. NWRank is less vulnerable to the TKC effect, and can find meaningful authorities where the mutual reinforcement approach (WRank and HITS algorithms) fails to do so.
The difference between rankings produced by different algorithms reflects the different features of each ranking algorithm. The most important feature of HITS and WRank algorithms is the mutual reinforcement of nodes or nodes and links, while the salient feature of PageRank is the weight normalization of the link. Edge betweenness approximately considers the role of network flow in the ranking. However, the NWRank algorithm combines the features of both mutual reinforcement from HITS and weight normalization from PageRank. In addition, NWRank also approximately considers the network flow in the simultaneous ranking of nodes and links, which is desirable in certain networks like infrastructure systems, where links carry commodities. Such link-centric information has not been traditionally featured by other ranking approaches. In addition, NWRank can obtain similar ranking results relative to the expensive N-1 removal based strategy, which is used as a benchmark for ranking comparison herein. However, NWRank is much more efficient than the traditional removal strategy and produces satisfactory results that emphasize integrity balancing local and global perspectives. NWRank tends to favor network integrity in a practical sense (favoring a large percentage of the network to be connected even if a small percentage is disconnected, which agrees with engineering design principles; whereas other rankings tend to favor elements that keep the entire network connected not admitting a suboptimal state). For instance, the link rankings for DT10 and SF20 show NWRank favoring edges that keep network integrity in a large percentage, whereas other rankings, including the N-1 approach based on efficiency favors links connecting to terminal nodes. Similar but more subtle trends are measured for the Grid25 and UG16 link rankings. NWRank prefers large partitions over large networks, which is desirable because NWRank considers network connectivity and network flow in the ranking of nodes and links.
To highlight the computational efficiency of NWRank, a relatively large transportation network of South Carolina with 1,486 nodes and 2,321 links is selected to test the computational requirements of NWRank. It takes 1,551 seconds to calculate the ranking results if a common N-1 removal strategy is [fig_ref] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks [/fig_ref] [fig_ref] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks [/fig_ref] [fig_ref] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks [/fig_ref] 15 to rank the network nodes and links. Results show that if α is large, network connectivity plays a great role which means that nodes with high degree and associated links rank high. If 1-α is large, approximate network flow has more effect on the ranking which means that the nodes and links with high betweenness values rank high. In addition, the difference between the ranking score of links decreases with the increase in the weight of α versus 1-α. Given the results of the sensitivity study, the selection of the weight coefficients should be conducted according to the decision makers' preference. If network connectivity is more important, relatively larger α values [e.g., (0.75, 0.25)] can be used. If network flow is more important, relatively smaller α values [e.g., (0.25, 0.75)] can be used. Otherwise, equal weight (i.e. α = 0.5) can be chosen as an unbiased point.
# Discussion
This study proposes a new mutually reinforcing network ranking algorithm (NWRank) which combines the mutual reinforcement feature of HITS and the weight normalization feature of PageRank and also considers the role of network flow approximately. This new algorithm is different in that it can rank the nodes and links of a network simultaneously, which is important for certain types of networks, particularly engineering infrastructure systems whose links are as critical as nodes. Numerical experiment results show that NWRank can obtain similar node ranking results relative to the HITS and WRank algorithms for networks without the TKC effect or prestigious nodes. For networks with TKC effects or prestigious nodes, NWRank can avoid the problem that exists in the HITS and WRank algorithms. For the node ranking, NWRank is somewhere in-between the rankings produced by HITS, WRank, eigenvector centrality, and PageRank, and performs better than the four algorithms because it combines the advantages of both HITS and PageRank algorithms while avoiding the TKC effect. For the link ranking, NWRank is better than WRank and edge betweenness based on the comparison with an N-1 expensive removal strategy, and it can capture features of both network connectivity and approximate network flow as shown with the networks in this study. NWRank enhances upon the current practical or state-of-the-art topological methods on network component ranking in terms of scope, computational complexity, and applicability. NWRank provides a new way to rank nodes and links of a network efficiently, which has practical applications in many types of networks and queries. For instance, to find cost-effective paths in supply chain networks, or to prioritize resource allocation for upgrade of infrastructure systems, such as transportation networks, water systems, and power grids among others, all with the aim to support decision making and design, particularly when balancing local connectivity and global flow. Also, if detailed information of the networks is available (e.g., link and node capacities, node supply and demand levels, and physics-based governing flow models), future work can compare approximate topological ranking strategies, as NWRank, with N-1 removal strategies measuring performance as capacitated networks that cannot violate physical constraints.
[fig] Figure 1: A low degree node that is important for network flow or its approximations, but missed in PageRank and HITS rankings. [/fig]
[fig] Figure 2: Topology of select hierarchical networks: (a) 8-node simple transportation network (T8); (b) 8-node "bat" network (B8) (c) 13-node three-cluster network (C13) which is symmetric under each link emanating from node 1, and (d) symmetric 15-node tree network (HT15). [/fig]
[fig] Figure 3: Topology of select distributed networks: (a) 25-node grid network (Grid 25); (b) 10-node Delaunay Triangulation (DT10); (c) 20-node Scale-Free distributed network (SF20); (d) 16-node uniform distributed network (UD16). Scientific RepoRts | 5:15141 | DOi: 10.1038/srep15141 [/fig]
[table] Table 2: Node and link ranking results for the T8 and B8 hierarchical networks. *W = WRank; NW = NWRank; P = PageRank; H = HITS; E = eigenvector centrality; B = edge betweenness; R = N-1 removal strategy. [/table]
[table] Table 4: Basic information of the selected distributed networks. *developed by the author and his collaborators. [/table]
[table] Table 5: Node and link ranking results of the Grid25 and DT10 distributed networks. [/table]
[table] Table 6: Node and link ranking results of the SF20 and UD16 distributed networks. [/table]
[table] Table 7: Comparison of the node ranking between the ranking algorithms and N-1 removal strategy based on cosine similarity with the boldface used to demonstrate the most similar ranking.Table 8. Comparison of the link ranking between the ranking algorithms and N-1 removal strategy based on cosine similarity with the boldface used to demonstrate the most similar ranking. RepoRts | 5:15141 | DOi: 10.1038/srep15141 [/table]
|
Cyclic tensile strain facilitates ossification of the cervical posterior longitudinal ligament via increased Indian hedgehog signaling
The pathomechanisms of initiation and progression of ossification of the posterior longitudinal ligament (OPLL) are unclear. Indian hedgehog (Ihh) and related signaling molecules are key factors in normal enchondral ossification. The purpose of this study is to investigate the contribution of mechanical strain to OPLL and the relationship of Ihh with OPLL. Sections of the posterior longitudinal ligament (PLL) were obtained from 49 patients with OPLL and from 7 patients without OPLL. Cultured PLL cells were subjected to 24 hours of cyclic tensile strain. To identify differentially expressed genes associated with cyclic tensile strain, microarray analysis was performed. Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis identified upregulation of various genes, particularly of the Hedgehog signaling pathway; Ihh and related genes had increased expression compared with controls after 24hour cyclic tensile strain. In immunoblotting analysis, Ihh, Runx2, Sox9, Gli2, Gli3, and smoothened (SMO) had significantly increased expression after 6-or 12-hour cyclic tensile strain. OPLL samples were strongly immunopositive for Ihh, Sox9, Runx2, Gli2, Gli3, and SMO in the ossification front of OPLL. These results suggest that cyclic tensile strain induces abnormal activation of Ihh and related signaling molecules, and this might be important in the ossification process in OPLL.Ossification of the posterior longitudinal ligament (OPLL) is characterized by pathological bone formation and has been extensively investigated using various approaches 1-3 . Ossification commences in the vertebral posterior longitudinal ligaments, with a particular predilection for the cervical area, but intensifies and spreads with progression of the disease, resulting in osseous projections and compression of the spinal cord 4 . Myelopathy symptoms due to OPLL are common in middle-aged patients, especially around age 50 years. This pathological entity is common in northeast Asians, and involvement of certain genes was identified in a recent genetic analysis 5-7 . Histologically, the ossifying extension process in OPLL resembles that of enchondral ossification in the growth plates of long bones 8,9 . Our group reported that the ossification process in OPLL and ossification of the ligamentum flavum (OLF) involve chondrocyte differentiation with unique expression of various transcriptional factors. Furthermore, we speculated that differentiation of hypertrophic chondrocytes around the calcified area of the ossification front was closely related to the ossification process 10,11 .During enchondral ossification of the growth cartilage, Indian hedgehog (Ihh) signaling plays an important role in differentiation and maturation of chondrocytes 12,13 . In our previous study, we suggested that Ihh signaling might also be important in the ossification process in OPLL, as well as in enchondral ossification. We found that expression levels for Ihh, parathyroid hormone-related peptide (PTHrP), and Sox9 were higher in cultured ligament cells derived from the posterior longitudinal ligament (PLL) of patients with OPLL than in those from the PLL of patients without OPLL. In addition, in an immunohistochemical study, we observed expression of Ihh in the ossification front in patients with OPLL 14 .Ossifying extension in OPLL may be influenced by genetics, endocrine disorders, and other local factors, including cyclic tensile strain 15-17 . The posterior longitudinal ligament is a two-layer structure that is subjected to distraction stress longitudinally and regulates spinal instability. The superficial layer is in close contact with open Scientific RepoRtS | (2020) 10:7231 | https://doi.org/10.1038/s41598-020-64304-w www.nature.com/scientificreports www.nature.com/scientificreports/ the dura mater and bridges 3 or 4 vertebra, whereas the deep layer is posterior to the vertebral body and connects two adjacent vertebra. Due to these anatomical features, the PLL has distraction strain along its longitudinal axis, and a large mechanical overload. This mechanical stress is important in the ossification process. Previously, we reported that cultured cells derived from ossified plaques of patients with OLF had upregulated mRNA expression for factors involved in the Wnt/β-catenin signaling pathway after 24-hour cyclic tensile strain 18 . In a recent clinical study in patients with cervical OPLL, cervical laminoplasty with instrumented fusion suppressed progression of OPLL in comparison with stand-alone laminoplasty 19,20. In this study, we used microarray analysis to identify activated signaling pathways in association with cyclic tensile strain. We focused on Ihh and genes in the Ihh signaling pathway, which are important in enchondral ossification of limbs, to determine if these factors are affected by cyclic tensile strain. Thus, the aim of the study is to investigate the effect of cyclic tensile strain on expression of Ihh and related signaling factors in cultured OPLL cells and to evaluate their localization in the ossification front immunohistochemically, with the goal of defining the relationship between Ihh and OPLL.ResultsA summary of demographic data of the patients in the study is given inTable 1.Cultured cell ALP activity and morphology/viability. To assess osteogenic differentiation, OPLL and non-OPLL cultured cells were stained ALP (red-purple stained cells)(Fig. 1A,B). Under non-stressed conditions, the ALP-positive rates were 48.2 ± 10.2% in OPLL cells and 17.6 ± 9.7% in non-OPLL cells(Fig. 1C). Morphologically, the cultured posterior longitudinal ligament cells exhibited a fibroblast-like, spindle-shaped appearance. The shapes of the cultured cells were almost the same in non-OPLL and OPLL samples(Fig. 1D,E). Under cyclic tensile strain, the viabilities of the attached non-OPLL and OPLL cells were 91.3 ± 3.08% and 92.1% ± 2.53%, respectively, in a 24-hour strain. At each time point, the viability of stressed OPLL cells did not differ significantly from that of non-OPLL cells(Fig. 1F,G).
## Identification of upregulated genes using gene ontology (go) and kyoto encyclopedia of genes and genomes (kegg) analysis.
Microarray analysis was performed to identify differentially expressed genes associated with OPLL after cyclic tensile strain. The significant upregulated genes in OPLL cultured cells compared to those in non-OPLL cells (difference of intensity >2.0) under non-stress condition and 24-hour cyclic tensile strain were identified by all GO terms in the biological process (see online) and all KEGG terms . Based on the results of the KEGG analysis, we found 7 candidate genes (Ihh, SMO, PTCH1, PTCH2, Gli2, Gli3, and STK36) from the Hedgehog signaling pathway among 27 genes that were significantly upregulated under 24-hour cyclic tensile strain. Changes in expression levels of Ihh and related genes in OPLL and non-OPLL cells shown as a ratio (24 hours/0 hour) are summarized in. Expression of Ihh, Runx2, Sox6, and vascular endothelin growth factor (VEGF) increased by >50% after cyclic tensile strain in all OPLL samples (case 1-3). Other genes with increased expression of >50% included Sox9, COL11A2, COL6A1, and bone morphogenetic protein (BMP2) in OPLL sample case 1, PTHrP in OPLL sample case 2, and COL11A2, patched 1 (PTCH1), Gli3, and serine/threonine protein kinase 36 (STK36) in OPLL sample case 3. In contrast, only Sox9 expression increased by >50% in non-OPLL sample case 1. www.nature.com/scientificreports www.nature.com/scientificreports/ immunoblot analysis. To confirm the expression of genes identified using microarrays, Ihh and related signaling factors were evaluated by immunoblot analysis. Changes in levels of Ihh, Runx2, Sox9, SMO, Gli2, and Gli3 after cyclic tensile strain in cultured OPLL cells and non-OPLL cells in immunoblot analysis are shown in. Under a non-stress condition, the expression levels of Ihh, Sox9, and SMO were significantly higher in OPLL cells than in non-OPLL cells. The expression levels for Ihh, Runx2, Sox9, Gli2, and Gli3 in OPLL cells increased significantly after 6-hour cyclic tensile strain. SMO increased significantly after 6-hour cyclic tensile strain, but decreased after 24 hours.
Histopathological and Immunohistochemical staining. To confirm the expression of genes identified using microarrays, Ihh and related signaling factors were evaluated by immunohistochemical analysis. In patients with OPLL, the ossification front had features with an irregular arrangement. The calcified cartilage layer and calcification front extended longitudinally to the fibrocartilage area. Around the calcification front, hypertrophic chondrocyte-like cells were present. In contrast, this layered structure was not observed in patients with non-OPLL, although a regular structure with a few chondrocytes was present. summarized the results of immunohistochemical analysis. In OPLL tissue, Ihh was strongly expressed in chondrocytes, particularly in proliferating chondrocytes in the fibrocartilage area near the calcification front. Proliferating chondrocytes in the fibrocartilage area had positive immunoreactivity for Runx2. Sox9 was expressed in proliferating chondrocytes, but not in hypertrophic chondrocytes. PTHrP was expressed around the calcification front in the calcified cartilage area. Gli2 was expressed in proliferating chondrocytes and cartilage matrix around the hypertrophic chondrocytes, and Gli3 was expressed in hypertrophic chondrocytes in the calcified cartilage area. In contrast, Ihh, Sox9, Runx2, PTHrP, Gli2, and Gli3 were not expressed in non-OPLL tissue.
# Discussion
OPLL is a multifactorial disease that involves genetic and environmental factors. Among the environmental factors, cyclic tensile strain has an important local effect on ossification in OPLL. Cyclic tensile strain is an important regulator of bone metabolism, and increases both the number of osteoblasts and the expression levels of various osteogenic genes, such as alkaline phosphatase (ALP), type I collagen, osteocalcin, and osteopontin. In a previous study, we showed that enlargement of the nucleus pulposus followed by herniation, disruption, and regenerative proliferation of cartilage in the annulus fibrous participated in initiation of calcification of the PLL in twy/twy mice. In addition, after cells of the protruded hyperplastic annulus fibrosus invaded the longitudinal ligaments, these cells induced metaplasia of primitive mesenchymal cells to osteoblasts with neovascularization.
Regarding genetic effects, there is a high prevalence of OPLL in certain countries and races, and several reports have identified underlying genetic factors. The previous study demonstrated that patients with OPLL had a significantly higher incidence of genetic abnormalities in the type XI collagen (alpha) 2 gene (COL11A2) region, and genome-wide linkage study showed that a single-nucleotide polymorphism in the collagen 6A1 gene (COL6A1) was strongly associated with OPLL. In a DNA microarray analysis, angiopoietin-1 levels in cultured OPLL cells were significantly higher after adding chondrogenic factors, such as connective tissue growth factors or cartilage oligomeric matrix protein. Another study identified six susceptibility gene loci for progression of OPLL at 20p12.3, 8q23.1, 12p11.22, 12p12.2, 8q23.3, and 6p21.1 in a genome-wide association study. In addition, we have found that expression levels for Ihh, PTHrP, and Sox9 are significantly higher in cultured cells from patients with OPLL than in cells from non-OPLL patients. In our study, ALP activities of cultured cells derived from OPLL patients were significant higher than the cells from non-OPLL patients. These findings indicated that cultured OPLL cells have both osteogenic and cartilaginous (osteochondral bi-potential) characters.
Cyclic tensile strain also influences upregulation of expression of genes involved in progression of OPLL. A previous study demonstrated that mechanical stress due to cyclic stretching led to upregulation of several genes associated with bone metabolism, including endothelin (ET)-1 and prostaglandin I 2 , in OPLL cells. Spinal ligament cells derived from patients with OPLL had osteoblastic phenotypes and higher sensitivity to mechanical stress, compared with spinal ligament cells from patients without OPLL. In the present study, many osteogenic and chondrogenic markers, including Ihh signaling, were upregulated by cyclic tensile strain in a microarray analysis although there were individual differences in the results of gene expression. These results suggest that expression of Ihh and its related signaling factors might be sensitive to cyclic tensile strain, and additional local cyclic tensile strain might induce abnormal changes in expression of these genes.
We reported previously that OLF cells activate their ossification properties under cyclic tensile strain, via the Wnt/β-catenin signaling pathway. In addition, the other study reported that mRNA levels for ALP, osteopontin, BMP2, BMP4, and BMP receptors were significantly increased by cyclic stretching in cultured cells derived from the ligamentum flavum of patients with OPLL. In the present study, the expression levels for receptors of Ihh, such as Gli2 and Gli3, increased significantly after 6-hour cyclic tensile strain in immunoblotting analysis. Interestingly, Gli2 and Gli3 were upregulated before Ihh, Son9, Runx2, and SMO were upregulated. In patients www.nature.com/scientificreports www.nature.com/scientificreports/ with OPLL, the PLL is subjected to strain stress for a long time, and ossification progresses very slowly; however, our results demonstrate that cyclic tensile strain might affect the receptors first, followed by upregulation of Ihh, Sox9, Runx2, and SMO by strain stress.
In normal enchondral ossification, Ihh, a member of the Hedgehog family, is an essential factor for chondrocyte differentiation. In the growth plate, Ihh is expressed in prehypertrophic chondrocytes and causes the surrounding tissue to differentiate into metaphasic tissues. Ihh with Sox9 or PTHrP induces differentiation and maturation of chondrocytes in the early stage of normal enchondral ossification, and Ihh induces calcification with Runx2 in the late stage by controlling maturation of chondrocytes. In this study, Ihh was expressed in proliferating chondrocytes in the ossification front of human cervical OPLL plaques. Therefore, we speculate that Ihh might be involved in the ossification process in human OPLL, as in normal enchondral ossification.
Previous immunohistochemical studies have shown that chondrocytes near the ossification front express transcriptional factors and collagen types that are involved in ossification. The previous study described that BMP2 contributed to cartilage and bone formation and that TGF-β stimulated bone formation in the ossified ligaments www.nature.com/scientificreports www.nature.com/scientificreports/ of patients with OPLL 37 . We previously reported that type I collagen was significantly expressed in the matrix of the ossified PLL, whereas type II collagen was highly expressed in metaplastic chondrocytes at the ossification front, and that apoptotic hypertrophic chondrocytes were observed near the calcification front. In the present study, we showed that Ihh, Runx2 and Sox9 are expressed in proliferating chondrocytes. Furthermore, we demonstrated expression of PTHrP, Gli2 and Gli3 in hypertrophic chondrocytes. The presence of Gli2, Gli3, and SMO in the ossification front provides evidence that Ihh and its related signaling factors might contribute to the ossification process in human OPLL.
In conclusion, Gli2 and Gli3 were strongly upregulated after 6-hour cyclic tensile strain followed by upregulation of Ihh, Sox9, Runx2, and SMO, and genes related to chondrogenesis, including Ihh, Sox9, and Runx2, were upregulated after 24-hour strain in cultured OPLL cells. In addition, Ihh, Runx2, and Sox9 proteins were expressed in the proliferating chondrocyte layer, PTHrP, Gli2 and Gli3were expressed in the hypertrophic chondrocyte layer at the ossification front of OPLL. These results suggest that cyclic tensile strain might induce upregulation of genes and proteins related to differentiation or maturation of chondrocytes, and that Ihh and its related signaling factors might play a key role in human OPLL.
The limitation of this study was that there were differences in results between the in vivo and in vitro studies using human OPLL-related samples. Thus, cellular responses in cultured cells cannot be extrapolated with certainty to the in vivo situation of cells located in the extracellular matrix. Also, the cyclic tensile strain used for cultured cells was applied over a shorter period to that in vivo. Thus, the effects of mechanical stress on expression of signal transduction factors in OPLL requires further study.
# Methods
Patients and biological samples. En bloc samples of the ligament-ossified plaque complex were obtained from 49 patients (29 men, 20 women; mean age 69.6 years) who underwent anterior decompressive surgery at our hospital. OPLL was diagnosed based on the presence of bone formation in the PLL on radiography or computed tomography of the cervical spine. None of the patients had evidence of congenital bone or joint disorders or musculoligamentous tissue abnormalities, and none had seronegative spondyloarthropathy or hyperparathyroidism or was being treated with glucocorticoids or immunosuppressants. The surgical procedures for cervical anterior decompression and interbody fusion are described in detail in our previous publication. Briefly, a left-sided anterolateral approach was used to expose the affected vertebral levels. The OPLL plaque was isolated circumferentially, together with the surrounding non-ossified PLL (deep layer), like a floating "island" on the dura mater. Using a custom-made micro-Kerrison rongeur, the non-ossified area of the ligament peripheral to the ossified lesion was carefully dissected and removed en bloc with the surrounding tissues. Samples were also obtained from 7 patients without OPLL (cervical disc herniation: CDH) who underwent cervical anterior decompressive surgery (4 men, 3 women; mean age 63.4 years) for use as age-matched control materials . Before surgery, written informed consent was obtained from each patient. The study protocol was approved by the Human Ethics Review Committee of our University Medical Faculty and strictly followed the Clinical Research Guidelines of the Ministry of Health, Labor, and Welfare of the Japanese Government.
Histopathological staining. Resected OPLL plaque together with the surrounding posterior longitudinal ligaments tissue was bisected mid-sagittally, followed by fixation with 10% buffered formaldehyde for 48 hours at 4 °C. The sample was further decalcified for 4 to 7 days at 4 °C in 0.5 M EDTA and 0.5 M Tris-HCl buffer at pH 7.6, and then embedded in paraffin using standard procedures. Serial 4-µm thick sagittal sections of the OPLL were prepared for hematoxylin and eosin (H&E) and Toluidine-Blue staining.
cell culture. For cell culture, PLL specimens were harvested from 9 of the 49 patients with OPLL (7 men, 2 women; mean age at surgery 68.4 years; range 59-79 years: OPLL cells). Samples were also obtained from 7 patients without OPLL who underwent cervical anterior decompressive surgery (4 men, 3 women; mean age 63.4 years; range 56-75 years: non-OPLL cells) for use as age-matched control materials . Cultured PLL cells derived from the vicinity of the OPLL plaque (OPLL cells) were isolated as described previously. Briefly, the ligaments were harvested aseptically from the ossified tissue, minced into approximately 0.5-mm 3 pieces, and explanted onto 100-mm diameter dishes in 10 ml of Dulbecco's Modified Eagle Medium (#12320, Low Glucose 1X, Lot 561521, GIBCO, Grand Island, NY) supplemented with 10% qualified fetal bovine serum (#1395965, GIBCO) and 1% penicillin/streptomycin (#15140, 667553, GIBCO) and incubated at 37 °C in a . Summary of immunohistochemical staining of the ossification front. ++, strongly positive; +, moderately positive; ±, weakly positive; −, negative staining OA, ossified area; CCA, calcified cartilage area; FCA, fibrocartilage area, FA, fiber area.
[formula] Ihh ± + ++ − Runx2 − − + ± Sox9 ± − ++ ± PTHrP − + ± − Gli2 ± + + − Gli3 − + ± − [/formula]
www.nature.com/scientificreports www.nature.com/scientificreports/ humidified atmosphere of 95% air/5% CO 2 . The cultures were left undisturbed for 48 h, followed by replacement of the medium with an equal volume of fresh medium. Cells derived from the explants were harvested from the dishes with 0.02% ethylenediaminetetraacetic acid and 0.05% trypsin for further passages, and used in experiments after the third passage. ALP activity. Alkaline phosphatase activity was evaluated using an ALP stain kit (Takara Bio Inc., Shiga, Japan). For each sample of non-stressed cultured non-OPLL and OPLL cells, a micrograph of 10 fields was obtained for analysis, and the ratio of positively stained cells to total cells was determined.
In vitro application of cyclic tensile strain to cultured cells. A Flexercell ® FX-3000 cell stretching device (Flexercell International, Hillsborough, NC) was used in this study. The device consists of a computer-controlled vacuum unit and a culture-well plate with a flexible-polystyrene culture bottom coated with type I collagen (Flex I, Bioflex ® Plates, Flexercell ® International). The culture plates are six-well (35-mm diameter) flexible bottomed culture plates with a hydrophilic surface. Application of a vacuum provides a hemispherically downward deformation force on the flexible bottom, resulting in a non-homogenous strain profile with a maximum at the periphery and a geometric decline toward zero at the center of the culture well bottom. Based on a previous study in which the same device was used, we set the tensile strain at a maximum of 20% elongation. Selection of this strain level was also based on the other previous studies. Panjabi and White showed that the elastic properties of the human spinal ligament are markedly altered when they are subjected to 26% elongation. The other study assessed 0-30% elongation and selected 20% elongation at 0.5 Hz because expression of ALP was reported to increase with an increasing extension ratio to reach a plateau at an extension ratio of 30%, and cell damage occurred once the extension ratio exceeded 30% 27 . To show the validity of this strain condition using in this study, cell morphology and viability were examined, as described below. For this experiment, OPLL cells were subjected to cycles of 10 s of 20% elongation and 10 s of relaxation for 6, 12, or 24 h. For microarray analysis, three OPLL cell lines were prepared without strain and with 24-hour strain, and two non-OPLL cell line was used as a control. For immunoblotting analysis, eight OPLL cell lines and six non-OPLL cell lines were prepared without strain and with 6-, 12-, and 24-hour strain.
## Quantification of cell survival. cell viability of cultured ligament cells was examined using a live/
Dead assay (Cambridge Biosciences, Cambridge, UK) under a confocal microscope (Leica TCS SP2, Wetzlar, Germany). Ethidium homodimer-1 cannot penetrate live cells but stains the DNA of dead cells red. Calcein-AM penetrates live cells, in which an esterase cleaves the molecule, which then fluoresces green, whereas dead cells do not contain the esterase. For visualization, the cells were soaked in the Live/Dead solution for 1 hour. The numbers of live (green) and dead (red) cells were counted manually. The cell survival rate was calculated.
Microarray analysis. For microarray analysis, 3 OPLL and 2 non-OPLL cells were used . After cyclic tensile stretching, the cultured cells were disrupted in a lysis buffer containing β-mercaptoethanol, and total RNA extracts from the cells were pooled at each time point and further purified using RNeasy ® (Qiagen, Valencia, CA). A quality check of each total RNA sample was carried out using an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA). Total RNA (100 µg) was then reverse-transcribed into double-stranded cDNA using oligo-dT primers. Biotinylated cRNA was synthesized using an in vitro transcription labeling kit (Agilent). Following fragmentation, 20 μg of cRNA was hybridized for 17 h at 65 °C on a Agilent SurePrint G3 Human GE 8x60K v2 Microarray. Gene chips were washed and scanned using an Agilent DNA Microarray Scanner (Agilent). The scanned images were quantified using Agilent Feature Extraction 7 software (Agilent), and the raw data for each hybridization were calculated. The obtained microarray data were initially processed using Gene Spring GX 12.0 software (Agilent). The ratio and differences in intensity between two corresponding genes on each array were calculated. A significant difference in the levels of gene expression was defined as a ratio of the difference of intensity of >2.0.
GO and KEGG analysis were performed for categorizing identified genes, and identifying specific genes and pathways by enrichment analysis. The Gene Ontology Consortium (www.geneontology.org/doc/GO.doc.html) maintains a controlled vocabulary database of functional descriptions for genes. In our study, we used the total list of GO terms within the biological process categories. The KEGG system (http://www.genome.jp/kegg/) provides a reference knowledge base for linking genomes to life through the process of pathway mapping. A list of genes in OPLL cultured cells with at least a 2.0-fold change in expression and a p-value <0.05 compared with those in non-OPLL cells were used for KEGG pathway analyses. immunoblot analysis. For immunoblot analysis, 8 OPLL and 6 non-OPLL cells were used . An immunoblot analysis method previously described by our laboratory was used to evaluate expression levels of transcriptional factors. Briefly, the sample was solubilized in RIPA buffer, homogenized, and then stored at −80 °C. The protein concentration was determined using a Bio-Rad DC protein assay kit (Lot 500-0116, Bio-Rad Laboratories, Hercules, CA). Total protein (80 µg/lane) was subjected to sodium dodecylsulfate polyacrylamide gel (15%) electrophoresis (SDS-PAGE) and transferred onto a polyvinylidene difluoride membrane in a semi-dry blot apparatus. The membranes were then washed twice in PBS and subsequently reacted with primary antibodies overnight at 4 °C and then with an anti-rabbit IgG antibody and avidin-biotinylated peroxidase complex (1:200; Envision ™ System-HRP Labeled Polymer, Dako) for 3 h. The primary antibodies used in this study were rabbit polyclonal anti-Ihh (lot 20450) (GeneTex, San Antonio, TX, USA), rabbit polyclonal anti-Sox9 (lot D406), rabbit polyclonal anti-Runx2 (lot G1009) (both Santa Cruz Biotechnology, Santa Cruz, CA, USA), rabbit polyclonal anti-Gli2 (lot 584061) (CosmoBio, Tokyo, Japan), rabbit polyclonal anti-Gli3 (Lot GR79855-4), and rabbit polyclonal anti-smoothened (SMO: lot GR198520-1) (both Abcam, Tokyo, Japan). After triple washing in 0.1 M PBS, the membrane was immersed and subjected to radiography to visualize the peroxidase activity and thus the level of antibody binding. To quantify the relative level of expression of the four evaluated transcriptional factors in Immunohistochemical staining. For immunohistochemical staining, PLL specimens were harvested from 40 of the 49 patients with OPLL (23 men, 17 women; mean age at surgery 69.9 years). Samples were also obtained from 3 patients without OPLL who underwent cervical anterior decompressive surgery (1 man, 2 women; mean age 60.7 years) for use as age-matched control materials . Tissue blocks were stained immunohistochemically, as described previously. The primary antibodies used in this study (sources given above, unless stated) were rabbit polyclonal anti-Ihh, rabbit polyclonal anti-Sox9, rabbit polyclonal anti-Runx2, mouse monoclonal anti-PTHrP (batch 17011; AbD Serotec, Oxford, UK), rabbit polyclonal anti-Gli2, and rabbit polyclonal anti-Gli3. Signals were detected by reaction with goat anti-mouse immunoglobulin antibodies conjugated to a peroxidase labeled-dextran polymer (EnVision TM , Peroxidase, Dako, Glostrup, Denmark) for monoclonal antibodies and with goat anti-rabbit immunoglobulin antibodies (EnVision + , Peroxidase, Dako) for polyclonal antibodies at 20 °C for 45 min, followed by rinsing with PBS at pH 7.4. To visualize the peroxidase color reaction, sections were incubated with DAB-HCl solution (CB090; Dojin Chemicals, Tokyo, 50 mg dissolved in 100 ml of 0.05 M Tris-HCl buffer at pH 7.4) at 20 °C for 10 min. Nuclear counterstaining was carried out with hematoxylin. Semiquantitative analysis conducted according to the method previously described. Briefly, we indicated the degree of immunoreactivity as: −not labeled or tissue was negative; ±, weak labeling above background or only traces of the tissue were stained; +, moderate labeling above background or less than one-third of the cells were stained; ++, intense labeling above background or more than one-third of the cells were stained.
# Statistical analysis.
All values are expressed as mean ± SD. Differences between groups were tested by one-way analysis of variance (ANOVA), with p < 0.05 in a Tukey post hoc analysis considered to indicate significance. For measurement of stained tissues, the inter-and intraobserver reliabilities were assessed by calculating intraclass correlation coefficients (ICC), with ICC values < ±0.75-1.00 considered to show excellent reliability. All statistical tests were performed using SPSS ver. 24.0 (SPSS, Chicago, IL).
## Data availability
Data generated and analyzed during this study are included in this published article. Data and materials are available from the corresponding author subject to reasonable request and subject to the ethical approvals in place and materials transfer agreements. |
Rational design of Striga hermonthica-specific seed germination inhibitors
# Introduction
Striga hermonthica, commonly known as "purple witchweed," is a root-parasitic plant and considered one of the major biotic constraints to food production in sub-Saharan Africa [bib_ref] Parasitic plants major problem to food crops, Musselman [/bib_ref] [bib_ref] Observations on the current status of Orobanche and Striga problems worldwide, Parker [/bib_ref] [bib_ref] Armed and dangerous, Pennisi [/bib_ref] [bib_ref] Parasitic weed incidence and related economic losses in rice in Africa, Rodenburg [/bib_ref]. The infestation of cereals, such as maize (Zea mays), pearl millet, sorghum, and upland rice by Striga results in enormous yield losses. It can lead to complete crop failure, imposing a significant burden on the food security of more than 300 million people and causing economic costs of around seven billion US dollars annually [bib_ref] Major heretofore intractable biotic constraints to African food security that may be..., Gressel [/bib_ref] [bib_ref] Breeding for Striga resistance in sorghum: exploitation of an intricate host-parasite biology, Ejeta [/bib_ref] [bib_ref] Farmers' perspectives on the biotic constraint of Striga hermonthica and its control..., Atera [/bib_ref]. It is estimated that 50 million hectares of arable lands in Africa are infested by Striga [bib_ref] Major heretofore intractable biotic constraints to African food security that may be..., Gressel [/bib_ref] [bib_ref] Breeding for Striga resistance in sorghum: exploitation of an intricate host-parasite biology, Ejeta [/bib_ref]. Striga is characterized by exceptional fertility that gives rise to thousands of seeds per plant with 10-15 years of longevity in soil [bib_ref] Breeding for Striga resistance in sorghum: exploitation of an intricate host-parasite biology, Ejeta [/bib_ref] [bib_ref] Striga hermonthica parasitism in maize in response to N and P fertilisers, Jamil [/bib_ref] , building up large seed reservoirs in infested fields and appearing as one of the main obstacles in the long-term management of Striga . Albeit the employment of a number of mechanical, cultural, biological, and chemical control measures by local farmers, Striga is still a major problem for agriculture in sub-Saharan Africa [bib_ref] Control of parasitic weeds, Eplee [/bib_ref] [bib_ref] Biology and management of weedy root parasites, Joel [/bib_ref] [bib_ref] Advanced technologies for parasitic weed control, Aly [/bib_ref] [bib_ref] Current progress in striga management, Jamil [/bib_ref]. Hence, there is an urgent demand for control strategies that can deplete accumulated seed bank in infested soils or specifically inhibit the germination of Striga seeds [bib_ref] Breeding for Striga resistance in sorghum: exploitation of an intricate host-parasite biology, Ejeta [/bib_ref] [bib_ref] Strigolactones and root infestation by plant-parasitic Striga, Orobanche and Phelipanche spp, Cardoso [/bib_ref] [bib_ref] Suicidal germination for parasitic weed control, Zwanenburg [/bib_ref] [bib_ref] Methyl phenlactonoates are efficient strigolactone analogs with simple structure, Jamil [/bib_ref] [bib_ref] Methylation at the C-3 0 in D-Ring of strigolactone analogs reduces biological..., Jamil [/bib_ref] [bib_ref] A new series of carlactonoic acid based strigolactone analogs for fundamental and..., Jamil [/bib_ref] [bib_ref] Suicidal germination as a control strategy for Striga hermonthica (Benth.) in smallholder..., Kountche [/bib_ref].
Like other root parasitic plants, Striga is an obligate parasite that needs a host to survive [bib_ref] The strigolactone story, Xie [/bib_ref]. To ensure that emerging seedlings have access to a suitable host, seeds of these weeds germinate only upon sensing host-released signaling molecules, mainly strigolactones (SLs; [bib_ref] Secondary metabolite signalling in host-parasitic plant interactions, Bouwmeester [/bib_ref] [bib_ref] Strigolactones, a novel carotenoid-derived plant hormone, Al-Babili [/bib_ref]. SLs are a group of carotenoid derived plant hormones consisting of a lactone ring (D-ring) coupled with an enol ether bridge to a second variable moiety [bib_ref] Plant apocarotenoids: from retrograde signaling to interspecific communication, Moreno [/bib_ref] [bib_ref] On the biosynthesis and evolution of apocarotenoid plant growth regulators, Wang [/bib_ref]. They are released into the rhizosphere to attract symbiotic mycorrhizal fungi and play pivotal roles in various aspects of plant development and response to biotic and abiotic stress [bib_ref] Strigolactone inhibition of shoot branching, Gomez-Roldan [/bib_ref] [bib_ref] Inhibition of shoot branching by new terpenoid plant hormones, Umehara [/bib_ref] [bib_ref] Strigolactone signaling is required for auxin-dependent stimulation of secondary growth in plants, Agusti [/bib_ref] [bib_ref] Strigolactones affect lateral root formation and root-hair elongation in Arabidopsis, Kapulnik [/bib_ref] [bib_ref] Physiological effects of the synthetic strigolactone analog GR24 on root system architecture..., Ruyter-Spira [/bib_ref] [bib_ref] Positive regulatory role of strigolactone in plant responses to drought and salt..., Ha [/bib_ref] [bib_ref] Apocarotenoids: old and new mediators of the arbuscular mycorrhizal symbiosis, Fiorilli [/bib_ref].
In Striga, the receptors responsible for SL-induced seed germination are evolutionarily derived from a KARRIKIN INSENSITIVE 2 (KAI2) receptor and have arisen through gene duplication and new functionalization replacing the karrikin ligand by SLs [bib_ref] Evidence that KARRIKIN-INSENSITIVE2 (KAI2) receptors may perceive an unknown signal that is..., Conn [/bib_ref] [bib_ref] Structural basis of unique ligand specificity of KAI2-like protein from parasitic weed..., Xu [/bib_ref]. Among the several Striga SL receptors, S. hermonthica HYPO-SENSITIVE TO LIGHT (ShHTLs), ShHTL7 is the most sensitive to SLs. It is assumed to play a crucial role in regulating Striga seed germination [bib_ref] Probing strigolactone receptors in Striga hermonthica with fluorescence, Tsuchiya [/bib_ref].
The dependency of germination on host-derived signals, such as SLs, opens up the possibility of combating root parasitic weeds by applying germination stimulants in the absence of a host [bib_ref] Practicality of the suicidal germination approach for controlling Striga hermonthica, Samejima [/bib_ref] [bib_ref] Methyl phenlactonoates are efficient strigolactone analogs with simple structure, Jamil [/bib_ref]. This "suicidal germination" strategy has been tested recently in field and showed very promising results in alleviating Striga infestation [bib_ref] Practicality of the suicidal germination approach for controlling Striga hermonthica, Samejima [/bib_ref] [bib_ref] Methyl phenlactonoates are efficient strigolactone analogs with simple structure, Jamil [/bib_ref] [bib_ref] A new series of carlactonoic acid based strigolactone analogs for fundamental and..., Jamil [/bib_ref] [bib_ref] Suicidal germination as a control strategy for Striga hermonthica (Benth.) in smallholder..., Kountche [/bib_ref]. In addition, the inhibition of SL receptors, that is, DWARF14 (D14) and KAI2-derived receptors, in parasitic and nonparasitic plants by specific SL antagonists is a promising approach for understanding SL perception, combating root parasitic plants, and regulating plant architecture [bib_ref] Effects of triazole derivatives on strigolactone levels and growth retardation in rice, Ito [/bib_ref] [bib_ref] Structural basis for specific inhibition of the highly sensitive ShHTL7 receptor, Hameed [/bib_ref] [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref]. Recently, we have elucidated the high-resolution atomic structure of ShHTL7 and discovered, by serendipity, that the common detergent Triton X-100 is a noncovalent, specific ligand of this receptor, which competes with the SL ligand and acts as SL antagonist . Several tests and modeling confirmed the specificity of Triton X-100 in binding to ShHTL7 among other Striga and nonparasitic plant SL receptors . In addition, the application of Triton X-100 reduced Striga seed germination in vitro without affecting seed germination or rice growth. These results support the idea of using Triton X-100 as a lead compound to design specific and more efficient Striga seed germination inhibitors for combating this weed . At the same time, [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref] reported the development of efficiently covalent inhibitors of D14, taking advantage of the nucleophilicity of the active site Ser residue that mediates SL hydrolysis. This nucleophilicity is generally known for serine hydrolases and enables covalent modification and, hence, covalent inhibition of these enzymes by using reactive electrophiles. Based on studies showing that N-heterocyclic urea structures targeting the serine in the active site are efficient inhibitors of serine hydrolases, [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref] prepared a set of triazole urea chemistries and tested their effect on D14 mediated SL perception and signal transduction. Activity tests and structure studies revealed the triazole urea compound KK094 as a potent, covalent inhibitor of the rice D14 [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref]. KK094 also inhibited the ShHTL7 receptor, however, with much lower efficiency, and reduced the rate of Striga seed germination [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref]. In the present work, we developed specific types of ShHTL7 inhibitors, KK023-N1 and -N2, by combining structural elements of Triton X-100 and triazole ureas. The developed compounds, particularly KK023-N1, showed promising activity in blocking ShHTL7 and inhibiting Striga seed germination. In contrast to Triton X-100 and KK094, the developed inhibitors did not show any negative effect on rice as host plant. These findings show that KK023-N1 is a promising Striga-specific seed germination inhibitor that can be used for combating Striga in African agriculture.
# Results
Design, synthesis, and structure of KK023-N1 and KK023-N2
Previously, the structure of ShHTL7, co-crystallized with Triton X-100, showed that the hydrophobic tail of the ligand, which includes the phenoxy group, interacts with several amino acids in the receptor cavity . These interactions provide the basis for the tight binding of Triton X-100 and its blocking/inhibition activity, together with those that take place with the polyethylene repeats at the outer rim of the ShHTL7 binding pocket. Building upon this work, we co-crystallized ShHTL7 with the triazole urea derivative KK007, after incubating them together, and soaked the obtained crystal in 1 mM Triton X-100 solution. Thus, we determined a crystal structure that showed KK007and Triton X-100 bound to ShHTL7, where KK007 was covalently linked to the catalytic Ser95 , A-C; Supplemental . This structure showed that Triton X-100 left enough space in the active site to accommodate small covalent ligands targeting Ser95, and, hence, suggested that the combination of both moieties might generate efficient and Strigaspecific inhibitors. Therefore, we attached the alkyl benzene moiety, which corresponds to the hydrophobic tail of Triton X-100 without the methyl groups, to the triazole urea scaffold, which led to the two compounds KK023-N1 and KK023-N2. Attempts to introduce the methyl groups present in Triton X-100 into these hybrid structures were not successful. The synthesis scheme of KK023N1 and -N2 is depicted in [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref] , and the chemical structures of KK094, Triton X-100, and the two isomers are in [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref]. Nuclear magnetic resonance (NMR) analysis of the two compounds is shown in Supplemental .
## Kk023-n1 and kk023-n2 are specific inhibitors for shhtl7
To test the activity of the two compounds, we performed competition assays using Yoshimulactone Green (YLG) as a SL analog that produces a fluorescence signal upon hydrolysis by the SL receptors D14 and ShHTLs [bib_ref] Probing strigolactone receptors in Striga hermonthica with fluorescence, Tsuchiya [/bib_ref]. ShHTL7 efficiently hydrolyzed YLG, which was detected as an increase of the green fluorescence over time (Supplemental [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref]. This activity was completed using SL analog GR24 with an IC 50 of 0.27 mM [fig_ref] Figure 3: Inhibition of ShHTL7/OsD14 by KK023-N1 and KK023-N2 [/fig_ref]. Then, we tested if KK023-N1 and N2 will competitively inhibit ShHTL7-mediated YLG hydrolysis. As a result, KK023-N1 and N2 inhibited ShHTL7 hydrolysis activity with an IC 50 of 1.78 and 2.15 mM, respectively [fig_ref] Figure 3: Inhibition of ShHTL7/OsD14 by KK023-N1 and KK023-N2 [/fig_ref] , B and C), and this inhibition was concentration-dependent. Under the same conditions, Triton X-100 and KK094 inhibited ShHTL7 activity with an IC 50 of 0.47 mM and 425 mM, respectively (Supplemental [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref] , B and C). Furthermore, to check for the specificity of KK023-N1 and KK023-N2, we also evaluated their effect on the rice (Oryza sativa) SL receptor D14 (OsD14) receptor. Interestingly, neither of the two isomers could inhibit the hydrolytic activity of OsD14 (Figure 3, E and F), whereas GR24 showed an IC 50 of 2.59 mM [fig_ref] Figure 3: Inhibition of ShHTL7/OsD14 by KK023-N1 and KK023-N2 [/fig_ref] , indicating the high specificity of the two inhibitors toward the Striga ShHTL7 receptor. Likewise, we tested if KK023-N1, the more potent inhibitor, inhibits the Striga KAI2-derived SL receptor ShHTL5 that hydrolyzes YLG (Supplemental [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref]. We observed only weak inhibition of the hydrolysis activity of purified ShHTL5, albeit the high sequence similarity to ShHTL7 [bib_ref] Structural analysis of HTL and D14 proteins reveals the basis for ligand..., Xu [/bib_ref]. Indeed, the application of KK023-N1 at a 50 mM concentration led to only a 20% reduction of ShHTL5 hydrolysis activity, indicating an IC 50 4 50 mM (Supplemental [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref].
The specificity of KK023-N1 toward ShHTL7 could be seen from in silico docking of KK023-N1 to ShHTL7, where it could bind noncovalently in the lowest-energy model between the helices a3-a4 at the entrance of the ShHTL7 binding pocket, primarily through hydrophobic interactions with Thr157 and Ile161 [fig_ref] Figure 4: Docking of KK023-N1 to ShHTL7 [/fig_ref] , A-D). This model provides a rationale for the loss of binding by KK023-N1 toward OsD14 and ShHTL5 due to the presence of bulky amino Crystal structure of KK007 and Triton X-100 bound to ShHTL7. A, Crystal structure of KK007 (green stick) and Triton X-100 (yellow stick) bound to ShHTL7 (gray) (B) 2FoFc omit map (gray mesh) shows KK007 (green stick) covalently bound to the active site of ShHTL7 S95 (gray stick) contoured at 1 r cutoff (C) 2FoFc omit map (gray mesh) of the bound Triton X-100 (yellow stick) contoured at 1 r cutoff. KK023-N1 decreased binding affinity between ShHTL7 and ShMAX2-CTH and increased ShHTL7 thermal stability
We then investigated whether KK023-N1 interferes with the binding of ShHTL7 to the downstream signal transduction component S. hermonthica MORE AXILLARY BRANCHING 2 (ShMAX2). For this purpose, we first investigated if the Cterminal helix of ShMAX2 (ShMAX2-CTH) is sufficient for binding to ShHTL7 in the presence of the SL analog GR24, based on the known fact that the C-terminal helix of D3 is sufficient for interaction with D14 in the presence of GR24 [bib_ref] Structural plasticity of D3-D14 ubiquitin ligase in strigolactone signalling, Shabek [/bib_ref]. Using Microscale thermophoresis (MST), we found that GR24 facilitates binding between ShHTL7 and ShMAX2-CTH with a K d value of $100 mM, and there was no binding in the absence of GR24 , A and B). To determine whether KK023-N1 inhibits this interaction, we performed the measurement using ShHTL7 pre-incubated with KK023N1, which resulted in a substantial reduction in the binding affinity (K d = 663 mM), indicating that KK023-N1 blocks the ShHTL7 binding pocket and thus prevents the SL-dependent interaction with ShMAX2 . Interaction between ShHTL7 and CTH of MAX2 in the presence of GR24 was further confirmed by GST pull-down. We found GST tagged CTH of ShMAX2 could pull down His-ShHTL7 in the presence of GR24, and KK023-N1 could inhibit such interaction; also, we have used Triton X-100 as control which confirmed that KK023-N1 is sufficient to inhibit ShHTL7 activity .
To further confirm KK023-N1 binding to ShHTL7, we measured the thermal stability of the receptor, using GR24 as a comparison. The melting temperature of ShHTL7 decreased from 50 C to 43 C in the presence of high concentrations of GR24 . In contrast, incubation with KK023-N1 did not affect the melting temperature of ShHTL7. Interestingly, the application of high KK023-N1 concentrations decreased the fluorescence of the SYPRO orange indicating an interaction and binding of this inhibitor to the protein in a concentration-dependent manner. Triton X-100 also showed the same pattern as observed for KK023-N1 with a reduction in the peak fluorescence, while incubation with KK094 did not change the fluorescence peak , B-D).
## Kk023-n1 and kk023-n2 inhibited striga seed germination
The in vitro studies with the receptor ShHTL7 indicated that KK023-N1 and KK023-N2 could inhibit Striga seed germination. To test this possibility, we applied KK023-N1, -N2, Triton X-100, and KK094 at varying concentrations along with GR24 (at 0.5 or 1.0 nM) on preconditioned Striga seeds (collected from Sudan). Both KK023-N1 and -N2 showed inhibition of Striga seed germination in a concentrationdependent manner. In general, KK023-N1 appeared more effective than KK023-N2 , A and B). Application of KK023-N1 between 20 and 40 mM concentrations led to around 22%-51% reduction compared to the GR24 treatment (at 0.5 or 1.0 nM). KK094 showed complete inhibition at 80 mM, and the same trend has been reported previously [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref]. However, Triton X-100 appeared weaker than KK023, showing a 20%-35% reduction in Striga seed germination with a 20 mM concentration. In another study, we also tested the inhibitory activity of KK023-N1 and -N2 on Striga seeds obtained from Kenya. We found considerable reduction (28%-100%, depending on applied concentrations) in Striga seed germination by both inhibitors compared to control (Supplemental [fig_ref] Figure 3: Inhibition of ShHTL7/OsD14 by KK023-N1 and KK023-N2 [/fig_ref].
## Kk023-n1 and kk023-n2 did not impact rice seeds growth or tillering
In addition, we have assessed KK023-N1 and -N2 impact on rice seed germination and seedling development. For this purpose, we applied various concentrations (up to 100 mM) of both compounds on imbibed rice seeds and allowed them to grow for 2 weeks. We did not detect any effect on seed germination , A and B). Furthermore, rice seedlings treated with KK023-N1 or KK023-N2 did not show any significant phenotypic difference to the control seedlings. Previously, [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref] demonstrated that KK094 inhibits SL perception by D14 and induces the emergence of second tillers by interrupting the SL signaling pathway. This raises the question of whether our compounds also interfere with D14 and promote bud outgrowth in rice. To test this hypothesis and the specificity of KK023 isomers toward Striga weeds, we firstly applied KK023-N1, Triton X-100, and KK094 (at 10, 50, and 100 mM) to wild-type (WT) plants. We did not observe any significant negative impact of all concentrations of KK023-N1 on plant height and rice tillering in WT, except at 100 mM concentration that led to a reduction in biomass , A, C, and E). However, the application of Triton X-100 at high concentrations caused a reduction in plant height and dry biomass. The negative impact of Triton X-100 on dry biomass was detected already at a concentration of 10 mM, Both KK023 and Triton X-100 did not affect tillering in WT rice. In contrast, the application of KK094 led to a significant increase in rice tillering and a reduction in plant height and dry biomass in WT plants, demonstrating the negative impact of this compound on the growth of the host plant rice , A, C, and E). To further validate the effect of KK023-N1, Triton X-100, and KK094 on SL perception and ultimately on rice tillering phenotype, we applied them in combination with 1.0 mM GR24 to seedlings of the SL deficient rice mutant d17 . KK023-N1 and Triton X-100 did not block the GR24 effect on rice tillers , which indicated that D14 does not bind these two compounds, and their SL antagonistic activity is specific to Striga seed germination, whereas KK094 blocked D14, resulting in increased tillering phenotype .
## Kk023-n1 reduced striga emergence in infested rice under greenhouse conditions
Next, we evaluated the effect of KK023-N1 on Striga emergence in artificially infested pots planted with rice as a host crop [fig_ref] Figure 10: Effect of KK023-N1 and Triton X-100 on Striga emergence in rice under... [/fig_ref]. Striga emergence was recorded 10 weeks after rice planting. On an average, we observed maximum Striga emergence (26 per pot) in the mock-treated pots, while pots treated with KK023-N1 at 10 and 100 mM, showed an average of 20 and 16 Striga plants per pot, respectively, which corresponds to a reduction of 24%-38% in Striga emergence [fig_ref] Figure 10: Effect of KK023-N1 and Triton X-100 on Striga emergence in rice under... [/fig_ref]. Also, the Triton X-100 Determination of K d by MST for the interaction between ShHTL7 and ShMAX2-CTH. A, The binding affinity between 20 nM of labeled ShHTL7 and ShMAX2-CTH was analyzed in the buffer and no direct interaction was detected between the two proteins in the absence of GR24. B, Addition of 20 mM of GR24 appeared very active to induce the ShHTL7-ShMAX2-CTH interaction. C, The binding affinity shifted upon the addition of KK023-N1 with a dissociation constant K d = 663.13 ± 25.7 mM, indicating that KK023-N1 is blocking the GR24-dependent interaction between ShHTL7-ShMAX2-CTH. Data are the means ± SD (n = 3). D, GST pull down of His-ShHTL7 by CTH of ShMAX2 in the presence of GR24, Triton X-100, and KK023-N1. treated pots (at 10 and 100 mM) exhibited a reduction in the Striga emergence ($13%-25%). However, the number of emerged Striga plants in Triton X-100 treated pots was higher as compared to the KK023-N1 treatment. Plant height was recorded at the end of the study, and application of KK023-N1 showed even a positive impact on plant height, as compared to mock and Triton X-100 treatments [fig_ref] Figure 10: Effect of KK023-N1 and Triton X-100 on Striga emergence in rice under... [/fig_ref].
# Discussion
The control of the noxious parasitic weed Striga has become a significant challenge for global food production, particularly in sub-Saharan Africa, despite the efforts deployed to combat it over the last decades. The persistence of the Striga problem compels scientists to develop alternatives and more efficient methods for combating this weed. Suicidal germination, a strategy that takes advantage of the Evaluation of ShHTL7 thermal stability and interaction with KK023-N1 and GR24 using differential scanning 945 fluorimetry (DSF). Melting temperature curve for ShHTL7 at varying concentrations of (A) KK023-N1 or (B) GR24 is assessed by DSF. Protein unfolding was detected by the changes of fluorescence of SYPRO Orange dye. The inverse peak minima inferred to the melting temperature. Data are means (n = 3) measured in parallel.
## Figure 7
Effect of Triton X-100, KK023-N1, KK023-N2, and KK094 on S. hermonthica seed germination. Preconditioned Striga seeds ($50-100 per disc) were first pretreated with the indicated concentrations of inhibitors for 24 h, and then mixture of each inhibitor at the indicated concentrations along with GR24 at (A) 0.5 nM and (B) 1 nM was applied for another 24 h. Germinated and nongerminated seeds were counted from each disc to calculate percentage germination. Data are means ± SE (n = 6 discs). The negative value on top of each bar is the percentage reduction of Striga seed germination over control. dependency of parasitic seeds on host-released germination stimulants to diminish accumulated seed banks, has been recently demonstrated to be efficient in alleviating Striga infestation [bib_ref] A femtomolar-range suicide germination stimulant for the parasitic plant Striga hermonthica, Uraguchi [/bib_ref] [bib_ref] Suicidal germination as a control strategy for Striga hermonthica (Benth.) in smallholder..., Kountche [/bib_ref]. However, this strategy can only be applied in the absence of the host, which means that farmers cannot plant their fields with cereals when cleaning off their soils. The development of specific inhibitors of Striga seed germination offers a complementary approach for combating Striga, which can be applied in parallel to host crop cultivation. In this work, we combined the structure of Triton X-100 and the scaffold of triazole urea compounds (KK series; [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref] to develop chemistries with high efficiency in blocking Striga seed germination, and which are suitable for field application. Our combinatory approach led to the two isomers, KK023-N1 and KK023-N2, as inhibitor candidates.
We first characterized the inhibitory activity of the two candidates using in vitro YLG competition assays with purified ShHTL7 and D14 protein. The fluorescent substrate YLG has been widely used to measure the hydrolysis activity of SL receptors. The IC 50 determined for ShHTL7 in the presence of KK023-N1 or KK023-N2 were 1.78 and 2.15 mM, respectively, which are much lower than that observed for OsD14. Additionally, both KK023-N1 and KK023-N2 did not influence the YLG hydrolysis by ShHTL5, despite the high sequence similarity with ShHTL7. Structure analysis revealed that ShHTL7 possesses a large ligand-binding pocket among other SL receptors in parasitic and nonparasitic plants [bib_ref] Structural analysis of HTL and D14 proteins reveals the basis for ligand..., Xu [/bib_ref]. The increase in the pocket volume is due to the movement of the aD1 helix outward from the entrance of the pocket. Also, the ShHTL7 binding pocket is composed of nonbulky residues, compared to OsD14 and ShHTL5, and this explained the basis for the ligand selectivity [bib_ref] Structural analysis of HTL and D14 proteins reveals the basis for ligand..., Xu [/bib_ref]. The presence of the triazole urea scaffold in KK023-N1 and -N2 makes covalent binding to the active site serine in ShHTL7 likely, as shown for KK007 and the D14 inhibitor KK094 [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref]. However, our YLG competition assay [fig_ref] Figure 3: Inhibition of ShHTL7/OsD14 by KK023-N1 and KK023-N2 [/fig_ref] , A-C) indicated that KK023-N1 is outcompeted by GR24, suggesting the binding to be noncovalent. Indeed, MALDI-TOF and LC-MS analysis of protein and peptides (Supplemental Method S1) obtained by digesting ShHTL7 upon incubation with KK023-N1 did not indicate covalent modification to the receptor (Supplemental [fig_ref] Figure 4: Docking of KK023-N1 to ShHTL7 [/fig_ref]. Our docking model of KK023-N1 bound to ShHTL7 also suggests that the compound does not reach deep enough in the ShHTL7 binding pocket to use its potential for covalent linkage with the catalytic Ser95 [fig_ref] Figure 4: Docking of KK023-N1 to ShHTL7 [/fig_ref]. Therefore, KK023-N1 might interact noncovalently with ShHTL7, as previously shown for Triton X-100 . Soporidine (SOP) Effect of KK023-N1 and KK023-N2 on rice seeds germination and seedlings growth. Rice seeds were germinated in 24-well plate in 0.5 MS medium along with various concentrations of (A and B) KK023-N1 and (C and D) KK023-N2. The growth of rice seedling was assessed for 1 week and no effect on plant height was observed under all treatments. Data are means ± SE (n = 6). Means not sharing a letter in common differ significantly at P 0.05 (analysis of variance (ANOVA) test). [bib_ref] Small-molecule antagonists of germination of the parasitic plant Striga hermonthica, Holbrook-Smith [/bib_ref] [bib_ref] Small-molecule inhibitors: weed-control measures, Yoneyama [/bib_ref] , a previously reported Striga seed germination inhibitor, was shown to competitively inhibit ShHTL7-mediated YLG hydrolysis with an IC 50 of 12.5 mM. This value is around six-fold higher than our inhibitors KK023-N1 and KK023-N2, indicating a higher efficiency of our compounds in blocking ShHTL7, the most sensitive SL receptor in Striga. Our YLG hydrolysis assay showed that KK023-N1 and KK023 could competitively prevent YLG from being hydrolyzed by ShHTL7, indicating that they are very likely bound to close proximity of the binding pocket and thus prevent YLG from being bound to the ShHTL7 active site. Moreover, our docking model also had the lowest binding energy possible when the compound is bound to the ShHTL7 pocket rather than any other site.
In SL signal transduction, the receptor D14 interacts with D3 (MAX2 homolog) in a SL-dependent manner [bib_ref] Specialisation within the DWARF14 protein family confers distinct responses to karrikins and..., Waters [/bib_ref]. Upon binding to SL, D14 undergoes a conformational change that facilitates the interaction with MAX2 [bib_ref] DWARF 53 acts as a repressor of strigolactone signalling in rice, Jiang [/bib_ref] [bib_ref] Molecular mechanism of strigolactone perception by DWARF14, Nakamura [/bib_ref] [bib_ref] Rice DWARF14 acts as an unconventional hormone receptor for strigolactone, Yao [/bib_ref] [bib_ref] Regulation of biosynthesis, perception, and functions of strigolactones for promoting arbuscular mycorrhizal..., Yoneyama [/bib_ref] [bib_ref] On the biosynthesis and evolution of apocarotenoid plant growth regulators, Wang [/bib_ref]. Based on the high sequence similarity between ShMAX2 and D3 receptor anchoring site, ShHTL7 binding to ShMAX2 was shown to be also SL-dependent [bib_ref] ShHTL7 is a non-canonical receptor for strigolactones in root parasitic weeds, Yao [/bib_ref] [bib_ref] Structural basis for specific inhibition of the highly sensitive ShHTL7 receptor, Hameed [/bib_ref] , and this interaction was precluded by the presence of the inhibitor Triton X-100 [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref]. D3-CTH of rice was already shown to be sufficient for triggering this interaction [bib_ref] Structural plasticity of D3-D14 ubiquitin ligase in strigolactone signalling, Shabek [/bib_ref]. Using MST, we showed in our study that ShHTL7 binding to ShMAX-CTH is SL-dependent, suggesting that ShHTL7 undergoes a conformational rearrangement upon SL binding, which establishes the association with ShMAX2-CTH. The K d value for this interaction indicated modest binding affinity with 100 mM, but this could be because we only used a shorter purified CTH of ShMAX2 fused to GST. Also, taking into consideration that we might be missing other sites of ShMAX2 involved in ShHTL7-SHMAX2 complex formation. However, our study aimed to validate the association of this peptide. We have shown that it binds in a SL-dependent manner and that the addition of KK023-N1 perturbed SL binding caused a substantial, six-fold reduction in the affinity of ShHTL7 toward ShMAX2-CTH , A-C). GST pull-down study further confirmed that KK023-N1 could inhibit ShHTL7 activity and thus prevents it from hydrolyzing its substrate and binding to the downstream regulator . The effect of KK023-N1 on ShHTL7-ShMAX2-CTH interaction clearly demonstrates that the occupation of ShHTL7 binding pocket by the inhibitor prevents the occurrence of the structural changes required to anchor ShMAX2 and antagonist the effect of GR24, the synthetic SL. Moreover, we have shown using differential scanning fluorimetry (DSF) that the Effect of KK023-N1, Triton X-100, and KK094 on growth of rice seedlings in Nipponbare (WT) and d17 under hydroponics conditions. Phenotype of (A) Nipponbare (WT) and (B) d17 rice seedlings treated with varying concentrations of each compound (±GR24 in d17) for 2 weeks. C, Rice seedling height in response to KK023-N1, Triton X-100, and KK094 application. All the treatments of KK023-N1 in WT and in d17 (±1 mM GR24) did not affect plant height while high concentrations of Triton X-100 led to dwarfness (D). Effect of KK023-N1, Triton X-100, and KK094 application on rice tillering. Both KK023-N1 and Triton X-100 had no effect on rice tillering in WT (1 per plant) while application of KK094 in WT and d17 (±1 mM GR24) led to production of second tiller. E, Total dry biomass of rice as affected by application of KK023-N1, Triton X-100, and KK094. Total dry biomass of rice (root + shoot) remained unaffected by lower doses of KK023-N1 (±1 mM GR24 in d17) while high doses of KK023-N1 in WT and all doses of Triton X-100 and KK094 (Both WT and d17) caused reduction in total dry biomass. Data are means ± SE (n = 4). Means not sharing a letter in common differ significantly at P 0.05 (ANOVA test). occupation of the receptor cavity by KK023-N1 stabilized the ShHTL7 structure and reduced its susceptibility to thermal denaturation, decreasing the amount of the hydrophobic peptides are required for the binding with SYPRO orange dye. Likewise, Triton X-100 in DSF assay showed the same trend, and with higher concentrations, the peak depth was lower than that for ShHTL7. While GR24 binding decreased the ShHTL7 melting temperature by 7 C, the interaction with KK094 did not change the melting peaks' pattern, which is consistent with the IC 50 obtained from the YLG study (525 mM; Supplemental [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref]. Based on all studies that were conducted here and in [bib_ref] Triazole ureas covalently bind to strigolactone receptor and antagonize strigolactone responses, Nakamura [/bib_ref] , KK094 weakly inhibited ShHTL7 compared with Triton X-100 and KK023 isomers.
Following the in vitro tests with purified receptors, we performed Striga seed germination inhibition studies (in the presence of germination stimulant) under lab conditions and detected a reasonable activity with both compounds. Application of KK023-N1 (at 20-40 mM) along with 0.5 or 1 nM GR24 reduces the germination of Striga seeds by around 50%, compared to the treatment with GR24 alone, while KK023-N2 at the same concentrations caused around 42% decrease in germination . KK094 showed50% inhibition of Striga seeds germination at 20 mM concentration and led to complete inhibition (100%) at 100 mM. Application of Triton X-100 (at 20-40 mM) showed a 31% reduction in Striga seed germination, which is below the values reported before . This difference in Triton X-100 activity might be attributed to the difference in Striga seeds batches used in each bioassay, that is, seeds used in this study showed higher germination activity than those used by [bib_ref] Structural basis for specific inhibition of the highly sensitive ShHTL7 receptor, Hameed [/bib_ref]. Moreover, the value obtained for the inhibitory effect of Triton X-100 on Striga seed was the outcome of the median of variant GR24 concentrations (1, 0.5, 0.25, and 0.125 nM) indicated by [bib_ref] Structural basis for specific inhibition of the highly sensitive ShHTL7 receptor, Hameed [/bib_ref].
We also assessed the effect of KK023-N1 and -N2 on the growth of rice as a host plant and did not observe any negative impact on plant height of the treated seedlings up to 100 mM . In another study, the application of KK023-N1 at 10, 50, and 100 mM did not show any effect on plant height of WT and SL deficient d17 (±GR24) rice seedlings . Application of KK023-N1 at 100 mM showed minor reduction in the biomass of rice seedlings, compared to KK094 or Triton X-100. Moreover, we did not detect a negative impact of KK023-N1 on rice in our greenhouse study. Our findings also show that Triton X-100 and KK023 isomers are ShHTL7 specific and selective inhibitors and did not bind to OsD14 in vitro. Here, we have provided strong evidence that none of these inhibitors (KK023-N1, Triton X-100) restore the tillering phenotype and do not interfere with SL perception in the host plants.
In addition, the application of KK023-N1 and -N2 on other Striga ecotypes (Kenya) reduced their germination rate. Consequently, these outcomes confirm the possible use of the proposed inhibitors (KK023-N1 and -N2) against Striga in host plant presence. Based on the results of YLG competition assays, Striga germination bioassays, and Striga pot study, KK023-N1 appeared more active than KK023-N2 with a 38% reduction in Striga emergence under greenhouse conditions [fig_ref] Figure 10: Effect of KK023-N1 and Triton X-100 on Striga emergence in rice under... [/fig_ref]. KK094 expressed a high inhibitory activity on Striga seeds germination compared to KK023 and Triton X-100. Yet, the negative effect of KK094 application on host plants (hydroponic study has to be overcome, and further modifications have to be done in order to use this compound as a specific inhibitor for Striga seed germination.
Inhibiting the Striga seed germination by blocking the perception of the SL germination stimulant could be a very efficient approach for combating Striga. Our hybrid inhibitors, particularly KK023-N1, showed potential inhibition in Striga seed germination and are characterized by their specificity and low impact on the host. These encouraging findings indicate the potential of these compounds for application in combating Striga and make them excellent lead compounds for developing further Striga-Specific herbicides that can be applied in the presence of a host. For practical field application, we are planning field trials in infested regions in Burkina Faso in the future.
# Materials and methods
## Chemicals and formulation
The standard SL analog GR24 was kindly provided by Prof. Binne Zwanenburg, Radboud University, Netherlands. GR24 was applied as a racemic mixture (rac-GR24) of two stereoisomers that differ in the configuration of C2 0 atom (2 0 R and 2 0 S configuration). Other chemicals used to prepare half-strength Hoagland's nutrient solutions, 2-(N-morpholino) ethane sulfonic acid (MES), and dimethyl sulfoxide (DMSO) were purchased from different suppliers. Seeds of S. hermonthica were collected from an infested Sorghum (Sorghum bicolor) field near Wad Medani, Sudan, and provided by Prof. Abdel Gabar Babiker. Striga seeds collected from Maize infested field from Kenya, were provided by Prof. Steven Runo (Kenyatta University, Kenya). Seeds of the Striga susceptible rice cv IAC 165 were kindly provided by Dr Jonne Rodenburg, Africa Rice, Tanzania.
Synthesis of 2-[3-(4-methoxyphenyl)propyl]-4,4,5,5tetramethyl-1,3,2 dioxaborolane A mixture of chloro(1,5-cyclooctadiene)iridium(I) dimer (90.6 mg, 0.135 mmol) and 1,2 Bis(diphenylphosphino)ethane (107.5 mg, 0.270 mmol) was stirred for 5 min at 40 C under argon atmosphere in dichloromethane (20 mL). Then 1-allyl-4-methoxybenzene (2.0 g, 13.49 mmol) and 4,4,5,5-tetramethyl-[1,3,2]dioxaborolane (2.1 g, 16.19 mmol) were added to the reaction mixture and stirred for 48 h at room temperature. After completion of the reaction, the mixture was passed through a silica gel short column to remove the unsolved material and then concentrated in vacuo. The residue was purified with silica gel (WakosilC-300HG) column chromatography (n-hexane:ethyl acetate) to give the target compound in 93% yield.
## Synthesis of b-[3-(4-methoxyphenyl)propyl]-boronic acid
A mixture of 2-[3-(4-methoxyphenyl)propyl]-4,4,5,5-tetramethyl-1,3,2-dioxaborolane (1.0 g, 3.62 mmol), acetone (43 mL), distilled water (21.7 mL), sodium periodate (2.32 g, 10.86 mmol), and ammonium acetate (837 mg, 10.86 mmol) was stirred for overnight at room temperature in a round bottom flask. The reaction mixture was then diluted with ethyl acetate (50 mL), washed with brine (50 mL). The aqueous phase was then extracted twice with ethyl acetate (30 mL). The combined organic phase was dried with sodium sulfate and concentrated in vacuo. The residue was purified with silica gel (WakosilC-300HG) column chromatography (n-hexane:ethyl acetate) to give the target compound in 72% yield.
Synthesis of the 4-Bromo-2,3-dihydro-indole-1-carboxylic acid tert-butyl ester A mixture of 4-Bromo-2,3-dihydro-1H-indole (1.0 g, 5.10 mmol), acetonitrile (10.2 mL), and Boc 2 O (1.34g, 6.12 mmol) was stirred overnight at room temperature in a round bottom flask. The reaction mixture was then diluted with organic solvent (50 mL of n-hexane:ethyl acetate) and washed with brine (50 mL). The combined organic phase was dried with sodium sulfate and concentrated in vacuo. The residue was purified with silica gel (WakosilC-300HG) column chromatography (n-hexane:ethyl acetate) to give the target compound in 98% yield.
## Synthesis of tert-butyl 4-(3-(4-methoxyphenyl)propyl)indoline-1-carboxylate
A mixture of B-[3-(4-methoxyphenyl)propyl]boronic acid(504.1 mg, 1.83 mmol), 2,3 dihydroindole-1-carboxylic acid tert-butyl ester (542.4 mg, 1.83 mmol), tetrahydrofuran (THF) (18.3 mL), potassium carbonate (759.7 mg, 5.49 mmol) and 1,1 0 -bis(diphenylphosphino)ferrocene-palladium(II)dichloride dichloromethane complex (137.1 mg, 0.165 mmol) was refluxed for 18 h. Then the reaction mixture was diluted with ethyl acetate (30 mL) and washed successively with water (50 mL) and brine (50 mL). The combined organic phase was dried with sodium sulfate and concentrated in vacuo. The residue was purified with silica gel (WakosilC-300HG) short column chromatography to remove the unsolved material and then concentrated in vacuo. The resulted oil was used in the next step without further purification.
## Synthesis of 4-(3-(4-methoxyphenyl)propyl)indoline
The mixture of tert-butyl 4-(3-(4-methoxyphenyl)propyl)indoline-1-carboxylate (1.83 mmol), ethyl acetate (18.0 mL) and hydrogen chloride ethyl acetate solution (4 mol/L, 1.0 mL) was stirred for 1 h at 60 C. The reaction mixture was concentrated in vacuo and purified with silica gel (WakosilC-300HG) column chromatography (n-hexane:ethyl acetate) to give the target compound in 67% yield.
## Synthesis of triton/triazole urea hybrids
A solution of 4-(3-(4-methoxyphenyl)propyl)indoline (252 mg, 0.94 mmol) and N,Ndiisopropylethylamine (DIPEA; 365 mg, 2.83 mmol) in THF (3 mL) was added dropwise to a triphosgene (140 mg, 0.47 mmol) solution in THF (5 mL) under ice-cooling. After stirring with cooling for 20 min, iced water (10 mL) and ethyl acetate (10 mL) were added to the reaction mixture, and extraction was carried out. The combined organic phase was washed with iced water, dried with anhydrous sodium sulfate, and concentrated in vacuo. The residue was dissolved in THF (5 mL). To this solution, 1H-1,2,3-Triazole (78 mg, 1.13 mmol), DIPEA (365 mg, 2.28 mmol), and DMAP (trace) were added at room temperature. After stirring at room temperature overnight, the mixture was diluted with ethyl acetate and washed with water. The organic phase was dried with anhydrous sodium sulfate, and the solvent was concentrated in vacuo. The residue was purified by silica gel column chromatography (n-hexane:ethyl acetate = 4:1-3:2) to give the products of N1isomer (141 mg, 41%) as a white solid and N2-isomer (72 mg, 21%) as a white solid.
Cloning, expression, and purification of ShHTL7, ShHTL5, OsD14, and ShMAX2CTH recombinant proteins cDNAs were cloned and proteins were purified as described earlier in [bib_ref] Structural basis for specific inhibition of the highly sensitive ShHTL7 receptor, Hameed [/bib_ref]. Briefly, ShHTL7, OsD14, and ShHTL5 were amplified using gene-specific primers (Supplemental and obtained polymerase chain reaction (PCR) fragments were digested with BamHI and XhoI for ShHTL7 and OsD14 or with EcoRI and XhoI for ShHTL5 and cloned into pGEX-6P-1 (GE Healthcare, Chicago, IL, USA) vector. For His-tagged ShHTL7, the cDNA was cloned between BamHI and NotI site to pQlinkH expression vector. Recombinant plasmids were transformed into Escherichia coli BL21 (DE3) and the Expression of the recombinant proteins was induced by the addition of 150-mM isopropyl b-D-1-thiogalactopyranoside, and induced cultures were incubated for 18 h at 16 C. Cells were harvested by centrifugation then resuspended in lysis buffer (50 mM Tris-HCl pH 8.0, 200 mM NaCl, 2 mM DTT, 0.5% (v/v) Tween-20) followed by sonication for 10 min with 2 s ON and 1 s OFF at amplitude 40% on ice. The Cleared lysate was loaded into the Glutathione Sepharose 4B resins and allowed to bind at 4 C for 2 h. For eluting the proteins, ShHTL7 and ShHTL5 were eluted by cleavage of GST tag using PreScission Protease (GE Healthcare), and OsD14 was eluted by adding 20 mM reduced glutathione (Sigma-Aldrich, St Louis, MO, USA) with adjusting the pH to 8.0. ShMAX2-CTH coding sequence was extracted using sequence alignment with reported CTH of OsD3 [bib_ref] Structural plasticity of D3-D14 ubiquitin ligase in strigolactone signalling, Shabek [/bib_ref] and synthesized as gblock (IDT) (Supplemental . ShMAX2-CTH was digested with EcoRI and XhoI and cloned into pGEX-6P-1, using CEPC cloning procedure [bib_ref] Circular polymerase extension cloning for high-throughput cloning of complex and combinatorial DNA..., Quan [/bib_ref]. Expression and purification of CTH were performed by adopting the same procedure as mentioned above. For His tagged ShHTL7 purification, it was lysed in the buffer (50-mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM DTT, 0.5% Tween-20) and allowed to bind Ni-NTA agarose beads (Qiagen, Hilden, Germany) for 1 h at 4 C. Then washed three times with 30 mM Imidazole for 15 min each wash and eluted with 250 mM Imidazole and then passed through gel filtration as above for other proteins.
Crystal structure determination of KK007 and Triton X-100 bound to ShHTL7
Purified ShHTL7 was incubated with an excess molar of KK007 and crystallized in buffer containing 20% (w/v) PEG 8000, 100 mM MES/Sodium hydroxide pH 6.0, and 200-mM calcium acetate under vapor diffusion method. Thus, the obtained crystal was soaked in crystallization buffer containing 1 mM Triton X-100. The crystals were flash cooled in liquid nitrogen with 25% (v/v) glycerol as cryo-protectant. Data were collected at 100 K at the beamlines Proxima 1 and Proxima 2A at the SOLEIL Synchrotron (France) . The data were processed in XDS. The structure was determined by molecular replacement using MoRDa with the ShHTL7 structure (PDB 5Z89) as a search model. The structures were manually investigated using Coot and refined using Phenix Refine (Supplemental . Co-ordinates for the crystal structure was deposited in PDB with accession code 7C8L.
## Ylg hydrolysis assays
About 0.33 mM of ShHTL7 protein was used to conduct in vitro hydrolysis of YLG (1 mM) as a substrate in a total volume of 100 mL of phosphate-buffered saline (1Â PBS) containing 0.1% DMSO in 96-well black plate (Greiner Kremsmünster, Austria; [bib_ref] Probing strigolactone receptors in Striga hermonthica with fluorescence, Tsuchiya [/bib_ref]. The hydrolysis of YLG over time and changes in the fluorescence intensity was measured by SpectraMax i3 (Molecular Devices, San Jose, CA, USA) using an excitation wavelength of 480 nm and detection wavelength of 520 nm and recorded at 10 min intervals for 2 h. YLG in 1Â PBS without protein was used as a control. For competitive inhibition of YLG hydrolysis in the presence of the compounds (KK023-N1 or -N2, Triton X-100, and KK094), proteins (ShHTL7, OsD14, ShHTL5, at 0.33 mM final concentration) were co-incubated with different concentrations ranges (20 nM to 50 mM) of both inhibitors for 1 h at room temperature, then 1 mM of YLG was added and incubated for another 1 h. Measurement of the fluorescence was conducted in SpectraMax i3 (Molecular Devices) using the above-mentioned method. The relative fluorescence unit was calculated by subtracting the auto-fluorescence of YLG in the buffer from the fluorescence values of YLG in the presence of a protein. The inhibitory curve and IC 50 were determined using https://www.aatbio.com/tools/ic50-calculator website.
MST measurement of the binding affinity of ShHTL7 to ShMAX2-CTH ShHTL7 was labeled with Alexa Fluro 488 TM (GREEN dye) in a 1:2 ratio. Then protein-dye mixture was passed through the Sephadex G-25 column to remove the unconjugated dye. The labeled protein (50 nM) was incubated with serial dilutions of ShMAX2-CTH in the absence or presence of 20 mM of GR24 (for 10 min), to determine the K d value of the interaction. In another experiment, 100 mM of KK023-N1 was coincubated with the labeled protein for 2 h, and 20 mM of GR24 was added for 10 min, followed by incubation with a serial dilution of ShMAX2-CTH. About 5 mL of the sample was loaded into premium monolith NT standard capillaries, and the assays were carried out in a NanoTemper Monolith NT.015T instrument. A gradient increase in the temperature was initiated using a laser diode, and the change in the excited fluorescence of the green dye was recorded. To determine the K d values of each reaction, the change in the thermophoresis of fluorescent was plotted against the concentrations of un-labeled ShMAX2-CTH, and data were analyzed by NanoTemper software .
## Gst pull-down of shhtl7 by cth of shmax2
Purified GST-CTH of ShMAX2 and GST were bound to the Glutathione Sepharose 4B Beads (GE healthcare). Purified 6 Â His-ShHTL7 was incubated with 100 lM KK023-N1 and Triton X-100 for 3 h, separately. Then apo ShHTL7 was added to the beads bound by GST-CTH of ShMAX2 in the presence and absence of 100 lM GR24. Similarly, ShHTL7 incubated with KK023-N1 and Triton X-100 were added to the beads bound by GST-CTH of ShMAX2 in the presence of 100 lM GR24. Apo ShHTL7 was added to the beads bound by GST only in the presence of 100 lM GR24 as a control. All the samples were washed with lysis buffer (50-mM Tris-HCl pH 8.0, 200 mM NaCl, 2 mM DTT) three times and then eluted with addition of 20 mM reduced glutathione. Thus, the eluted protein was subjected to Western blot using an antibody specific to 6XHis (Abcam, Cambridge, UK).
## Dsf assay
About 10 mM of ShHTL7 purified protein was incubated with various concentrations of KK023-N1, Triton X-100, or KK094 (ranging from 0 to 100 mM) for 30 min. Then SYPRO Orange (reporter dye) was added to 1Â final concentration. The reactions were carried out in 25 mL 1Â PBs buffer. The assay was performed on CFX96 Real-Time System (BIO-RAD, Hercules, CA, USA) thermal cycler on 96-well quantitative polymerase chain reaction (qPCR) white plate, and the plate was sealed with adhesive qPCR plate seal to prevent samples evaporation [bib_ref] Isothermal analysis of thermofluor data can readily provide quantitative binding affinities, Bai [/bib_ref]. First, samples were heated to 25 C for 1 min followed by linear increase in the temperature ramp from 25 C to 99 C at a rate of 0.5 C/15 s. The change in the protein folding was monitored by the changes in the fluorescence of the reporter dye, and the data were analyzed by the machine software that plots the inflection point of fluorescence against temperature. Data were exported into Excel, and replicates were averaged to generate the graph.
In silico docking of KK023-N1 to ShHTL7
KK023-N1 was docked to ShHTL7 using online server SwissDock server [bib_ref] SwissDock, a protein-small molecule docking web service based on EADock DSS, Grosdidier [/bib_ref]. Output results were analyzed to identify the best fitted model with low Gibb's free energy and biologically relevant structure. The output model was visualized using Pymol, and the final figures were made out of it.
## Striga seed germination bioassays
Striga seeds were surface sterilized by 20% Sodium hypochlorite (NaClO) for 10 min followed by successive 6 times washing with sterilized MilliQ water. Seeds were kept to dry under the laminar flow cabinet, and $50-100 sterilized Striga seeds were spread uniformly on a glass fiber filter paper disc (9 mm). For preconditioning, around 12 discs were put in a petri plate on a filter paper moistened with 3 mL sterilized MilliQ water. Plates were then sealed with parafilm and kept in an incubator for 10 d at 30 C (in dark). After preconditioning, plates were removed from the incubator and dried, and discs were pre-treated with different concentrations , and 640 lM) of KK023-N1, KK023-N2, Triton X-100, or KK094 for 24 h. Next day, a combination of the inhibitors (same concentrations) along with GR24 (0.5 nM or 1 nM) was applied, and discs were incubated at 30 C for 24 h. For inhibitory effect of KK023-N1 and -N2 on Striga seeds obtained from Kenya, seeds were treated with . After image scanning, germinated and nongerminated seeds were counted and germination rate (%) was calculated by using SeedQuant software [bib_ref] SeedQuant: a deep learning-based tool for assessing stimulant and inhibitor activity on..., Braguy [/bib_ref].
## Rice germination and growth in response to kk023-n inhibitors
Rice seeds (cv Nipponbare) were surface sterilized in 2.5% NaClO and 0.05% of Tween-20 for 10 min, and rinsed 6 times with sterilized MilliQ water. Seeds were imbibed in sterilized water at 30 C for 24 h in the dark. Next day, $3 mL of each inhibitor at different concentrations (3, 6, 12, 25, 50, and 100 lM) was added along with 0.5 Murashige and Skoog (MS) medium into 24-well plate, and one imbibed rice seed was placed in each well. The experiment was performed in six replications (one plant/well, n = 6) for each concentration. Nontreated rice seeds in 0.5-MS solution were included as a control. Seeds were allowed to germinate at 30 C in the dark for 48 h for monitoring seeds germination, and seeds were moved to a growth chamber with white fluorescent light (130-180 lM m -2 s -1 ) and12 h:12 h (Light/Dark (L/D)) period at 28 C. After 1 week, length of rice seedlings was measured and recorded, and pictures of seedlings were taken.
## Rice micro tillering assay
Rice seeds of WT (cv Nipponbare) or SL-deficient mutant (d17), were surface sterilized with 2.5% NaClO and 0.05% of Tween-20 for 10 min, and rinsed 6 times with sterilized MilliQ water. About 30 seeds were placed on sterilized Magenta GA-7 plant culture box containing 0.5% MS medium with 0.4% agar and incubated at 30 C in dark. After 2 d, boxes were moved to a growth chamber and allowed to grow for another 5 d under fluorescent white light (130-180 lM m -2 s -1 ). Equally grown seedlings were transplanted in 50-mL black tubes filled with Hoagland nutrient solution (1 seedling/tube). Seedlings were treated with 10, 50, and 100 mM of each compound (KK023-N1, Triton X-100, and KK094). The compounds were applied twice a week for 2 weeks to test their effects on the growth of WT plants. SL-deficient seedlings (d17) were treated as WT but with the addition of 1 mM GR24. Number of tillers per plant, plant height, and dry biomass were measured at final harvest.
## Striga emergence study under greenhouse conditions
Approximately 6,000 Striga seeds (20 mg) were weighed, mixed with sand/soil mixture (1:2 ratio), and put into a 3-L perforated plastic pot containing 0.5-L clean sand/ soil mixture in the bottom. The pots were kept under greenhouse-controlled conditions at 30 C and kept moist with water for Striga preconditioning for 10 d. Formulated KK023-N1 and Triton X-100 were applied (at 100 mL per pot) at 10 and 100 mM concentration. Then three rice seedlings (cv IAC165, 10 d old) were planted in each pot after 24 h of inhibitor application. All the treatments were applied continuously twice per week for 4 weeks. Water with same amount of solvent (Cyclohexanone + Atlas-G) was used as a control. Pots were irrigated with 200 mL of Hoagland's nutrient solution with low phosphate (4 mM) when needed. The rice plants were allowed to grow under greenhouse conditions for 10 weeks and the emerged Striga plants were counted from each pot. The height of rice plants was also measured to see impact of Striga infestation on host growth.
## Accession numbers
## Supplemental data
The following materials are available in the online version of this article.
Supplemental . NMR analysis of KK023-N1 and KK023-N2.
Supplemental [fig_ref] Figure 2: Synthesis of KK023-N1 and KK023-N2 [/fig_ref]. ShHTL7/ShHTL5-mediated YLG hydrolysis.
Supplemental [fig_ref] Figure 3: Inhibition of ShHTL7/OsD14 by KK023-N1 and KK023-N2 [/fig_ref]. Effect of KK023-N1 and KK023-N2 on S. hermonthica seed germination (Kenya batch).
Supplemental [fig_ref] Figure 4: Docking of KK023-N1 to ShHTL7 [/fig_ref]. Mass spectrometer analysis. Supplemental Method S1. MALDI-TOF and Nano LC-MS.
Supplemental . Data collection and refinement. Supplemental . Primer sequences used for cloning ShHTL7/OsD14 and ShHTL5 and ShMAX2-CTH gBlok sequences.
[fig] Figure 2: Synthesis of KK023-N1 and KK023-N2. A, The synthesis scheme (steps 1-6) of Triton X-100/1,2,3-triazole urea-based inhibitors. B, Chemical structures of KK094, Triton X-100, KK023-N1, and KK023-N2 isomers. [/fig]
[fig] Figure 3: Inhibition of ShHTL7/OsD14 by KK023-N1 and KK023-N2. A, concentration-dependent inhibition of ShHTL7-mediated YLG hydrolysis by (A) synthetic SL GR24 or by (B) KK023-N1 and (C) KK023-N2. Inhibition of OsD14-mediated YLG hydrolysis by (D) GR24, but not with KK023 isomers (E and F). Data are the means ± SD (n = 3).acids such as Tyr and Trp in their ligand cavity, which hinders the binding of KK023-N1. [/fig]
[fig] Figure 4: Docking of KK023-N1 to ShHTL7. A, Cartoon representation of KK023-N1 (green stick) docked to ShHTL7 (gray) (B) Surface representation of KK023-N1 (green sphere) docked to ShHTL7 (gray) (C) Ligplot shows residual contacts of ShHTL7 to docked KK023-N1 (D) Contact residues (gray) of ShHTL7 to docked KK023-N1 (green stick). [/fig]
[fig] Figure 10: Effect of KK023-N1 and Triton X-100 on Striga emergence in rice under greenhouse conditions. Formulated KK023-N1 or Triton X-100 was applied at two concentrations (10 and 100 mM), twice per week up to 4 weeks period. The emergence of Striga plants was recorded after 10 weeks of rice planting. A, Striga infestation in treated and nontreated rice plants. B, Number of Striga emergence in response to KK023-N1 and Triton X-100 application. The negative values on the top of each bar represent reduction in Striga emergence (in percentage) as compared to mock (C) rice plant height in treated and nontreated pots. Data are means ± SE (n = 4). [/fig]
|
Detection of the progression of retinal nerve fiber layer loss by optical coherence tomography in a patient with glaucomatocyclitic crisis
a b s t r a c tA 30-year-old man presented with glaucomatocyclitic crisis, also known as PosnereSchlossman syndrome, and acute intraocular pressure elevation. Changes in the longitudinal retinal nerve fiber layer (RNFL) and the cup-to-disc (C/D) ratio of the optic disc were detected by Stratus optical coherence tomography (OCT). The average RNFL thickness in the affected eye was greater than that in the unaffected eye on Day 3 (132.99 mm, C/D ratio: 0.24 vs. 105 mm, C/D ratio: 0.26). However, the RNFL thickness continued to decrease and the C/D ratio progressively increased in the affected eye over 12 months (60 mm, C/D ratio: 0.67), although the intraocular pressure was controlled at <21 mmHg during that period. Glaucomatous visual field defects were also found. A progressive decrease in the thickness of the RNFL was demonstrated by Stratus optical coherence tomography following an elevation in the acute intraocular pressure in the patient. Prompt treatment and longitudinal monitoring are necessary to prevent and detect glaucomatous damage.
# Introduction
Glaucomatocyclitic crisis, also known as PosnereSchlossman syndrome, is a unilateral, benign ocular disease that is characterized by a mild anterior chamber reaction accompanied by an increase in intraocular pressure (IOP) with open angles. It is often self-limiting and can be recurrent. However, the etiology is controversial. 1,2 Some patients have been known to develop glaucoma-related visual field defects as a result of repeated episodes of glaucomatocyclitic crisis. [bib_ref] Is Posner Schlossman syndrome benign?, Jap [/bib_ref] [bib_ref] PosnereSchlossman syndrome glaucomatocyclitic crisis, Shazly [/bib_ref] Using Stratus optical coherence tomography (OCT), we conducted a longitudinal evaluation of the thickness of the retinal nerve fiber layer (RNFL) and the cup-to-disc (C/D) area ratio of the optic disc in a patient with glaucomatocyclitic crisis. We examined the differences in these parameters at the acute and remission stages of the disease.
## Case report
A 30-year-old man presented with decreased vision in his right eye (OD) for 2 days. He had no history of systemic disease but had experienced a similar episode (OD) 1 year previously. During the first episode, the patient received treatment and had since been in a stable condition. Ocular examination of the OD revealed mild epithelial edema of the cornea, a few fine keratic precipitates, few aqueous cells, and little flare. The IOP was 46 mmHg with a hypopigmented open angle. The vitreous was clear, the retina was normal, and the optic disc exhibited mild hyperemia. The left eye (OS) had no abnormalities, and the IOP was 15 mmHg. The bestcorrected visual acuity was 20/20 (OU) with myopia of À0.5 diopters (OU). Color sensation and pupil response were normal (OU).
All laboratory data were normal, including antinuclear antibodies, rheumatoid factor, venereal disease research laboratory test, HLA-B27, and chest radiograph. Therefore, the patient was diagnosed with glaucomatocyclitic crisis. Following treatment with oral acetazolamide, topical 2% carteolol, and 1% prednisolone acetate solution, the IOP decreased to 12 mmHg (OD). The IOP was maintained at <15 mmHg (OU) with topical 2% carteolol solution (OD). Follow up was performed every month over a period of 12 months. The patient did not experience any repeat attacks during the 12-month follow-up period.
The thickness of the peripapillary RNFL and the C/D ratio were automatically calculated using the fast RNFL thickness protocol , and the optic disc parameters using a Stratus OCT 3000 (Version 4.0.2; Carl Zeiss Meditec, Inc., Dublin, CA, USA). Highquality images with a quality score of signal strength 10 (maximum 10) were obtained on Day 3, and at 2 months, 5 months, and 12 months after the episode of glaucomatocyclitic crisis. [bib_ref] Signal strength is an important determinant of accuracy of nerve fiber thickness..., Wu [/bib_ref] The average RNFL thickness in the affected eye (132.99 mm, C/D ratio: 0.24) was greater than that in the unaffected eye (105 mm, C/D ratio: 0.26) on Day 3. However, the RNFL thickness progressively decreased and the C/D ratio progressively increased at 2 months (73 mm, C/D ratio: 0.61), 5 months (67 mm, C/D ratio: 0.63), and 12 months (60 mm, C/D ratio: 0.67) (Figs. 1 and 2). By contrast, there were no obvious changes in the unaffected eye after 12 months (102 mm, C/D ratio: 0.25).
An automated perimetry test was conducted using the Humphrey 30-2 threshold program. After an episode of glaucomatocyclitic crisis in the affected eye, a glaucomatous superior arcuate visual field defect was found at initial examination and at 1 year (mean deviation: À7.22 dB, pattern standard deviation: 5.23 dB, left, vs. mean deviation: À7.28 dB, pattern standard deviation: 8.02 dB, right; [fig_ref] Figure 3: Superior arcuate visual field defect, initially [/fig_ref].
# Discussion
Investigations into the "normal" RNFL thickness, using Stratus OCT, revealed that between the ages of 18 years and 85 years, Caucasians have an average RNFL thickness of 98.1 ± 10.9 mm, Hispanics 103.7 ± 11.6 mm, and Asians 105.8 ± 9.2 mm. The mean RNFL thickness of the inferior quadrant is 126.1 ± 17.8 mm, superior quadrant 124.2 ± 17.9 mm, nasal quadrant 80.9 ± 18.1 mm, and temporal quadrant 69.0 ± 12.7 mm. The inferior and superior quadrants are thicker than the nasal and temporal quadrants. [bib_ref] Determinants of normal retinal nerve fiber layer thickness measured by Stratus OCT, Budenz [/bib_ref] Decreases in the RNFL thickness of the inferior and superior quadrants are more sensitive for the detection of glaucoma. [bib_ref] Optic nerve head and retinal nerve fiber layer analysis: a report by..., Lin [/bib_ref] A previous study demonstrated that after a single episode of acute primary angle closure with a high IOP (>50 mmHg), there was an initial increase, followed by a decrease, in the diffuse RNFL thickness. The average decrease in the RNFL thickness correlated with the interval of follow-up, and gradually stabilized at 3 months. The overall reduction in RNFL thickness correlated with the duration of acute IOP elevation. [bib_ref] Longitudinal changes in retinal nerve fiber layer thickness after acute primary angle..., Tsai [/bib_ref] In this case, most reduction in the thickness of the RNFL occurred within the first 2 months, with only a small reduction occurring after 2 months. This result is consistent with the findings of previous reports. [bib_ref] Longitudinal changes in retinal nerve fiber layer thickness after acute primary angle..., Tsai [/bib_ref] [bib_ref] Measurement of retinal nerve fiber layer in primary acute angle closure glaucoma..., Fang [/bib_ref] A possible explanation for this occurrence has been proposed by Yoles and Schwart. [bib_ref] Potential neuroprotective therapy for glaucomatous optic neuropathy, Yoles [/bib_ref] These authors proposed a mechanism, termed secondary degeneration, whereby glaucomatous neuropathy continues to progress even after the reduction of the high IOP. Furthermore, there may be an effect on the healthy neurons that were not part of the primary injury but were adjacent to the injured neurons, as they have been exposed to the degenerative environment.
The patient in our case had a relatively greater average RNFL thickness in the affected eye than in the unaffected eye on Day 3. The optic nerve head appeared slightly edematous after the high IOP persisted for up to 2 days prior to treatment, even after there was a reduction of the IOP in the acute stage. However, the visual field examination may be unreliable in the acute IOP elevation stage. After remission, the visual field defects vary in severity. [bib_ref] Perimetric defects after single acute angle-closure glaucoma attack, Bonomi [/bib_ref] This may explain why the visual field defect and the reduction of the RNFL thickness were not the same in the acute stage. Fang et al [bib_ref] Measurement of retinal nerve fiber layer in primary acute angle closure glaucoma..., Fang [/bib_ref] reported that most eyes affected by acute primary angle closure demonstrated an increase in the RNFL thickness within 2 weeks after the episode, except in one patient who had previously suffered acute attacks. The authors suggested that if the increase of the RNFL thickness due to edema is less than the decrease caused by glaucoma, RNFL atrophy may not be evident. Thus, the RNFL thickness may be in the normal range in the early period after an acute episode, especially in patients who experienced elevated IOP with ongoing RNFL damage prior to the acute attack.
After the episode, the decrease in thickness of the RNFL was generalized. However, the visual field defect, which occurred predominantly in the superior hemifield, existed prior to the episode, with only focal deepening after the episode. The possible explanation for the structureefunction discrepancy is that there was pre-existing RNFL thinning in the right eye prior to the episode, which was masked by the RNFL swelling that was caused by the attack. [bib_ref] Measurement of retinal nerve fiber layer in primary acute angle closure glaucoma..., Fang [/bib_ref] Jap et al 3 reported that 26.4% of patients developed glaucoma as a result of repeated attacks of glaucomatocyclitic crisis. Patients who experienced glaucomatocyclitic crisis for 10 years or more had a 2.8 times higher risk of developing glaucoma than those who had experienced the disease for less than 10 years.
Although glaucomatocyclitic crisis is a relatively benign and often self-limiting disease, progressive RNFL loss and glaucomatous damage can occur as a result of repeated attacks and prolonged IOP elevation, if there is no immediate treatment.
In conclusion, Stratus OCT demonstrated progression of RNFL thickness loss and C/D ratio enlargement after acute IOP elevation in a patient with glaucomatocyclitic crisis. Therefore, urgent treatment and longitudinal monitoring are necessary to prevent and detect glaucomatous damage in glaucomatocyclitic crisis. . Optic disc and retinal photographs at 12 months, after an episode of glaucomatocyclitic crisis in the affected eye. The optic disc cup/disc area ratio measured using Stratus optical coherence tomography was 0.67.
[fig] Figure 1: RNFL thickness, measured using Stratus optical coherence tomography, after an episode of glaucomatocyclitic crisis. The Avg.Thick, Savg, and Iavg were, respectively, 132.99 mm, 168 mm, and 154 mm in the affected eye (OD) on Day 3 (left); 67.3 mm, 79 mm, and 75 mm (OD) at 5 months (middle); and 60.33 mm, 76 mm, and 61 mm at 12 months (right). Avg.Thick ¼ average retinal nerve fiber layer thickness; Iavg ¼ retinal nerve fiber layer thickness of the inferior quadrant; INF ¼ inferior quadrant; NAS ¼ nasal quadrant; RNFL ¼ retinal nerve fiber layer; Savg ¼ retinal nerve fiber layer thickness of the superior quadrant; SUP ¼ superior quadrant; TEMP ¼ temporal quadrant. [/fig]
[fig] Figure 3: Superior arcuate visual field defect, initially (mean deviation: À7.22 dB, PSD: 5.23 dB, left) and at 1 year (mean deviation: À7.28 dB, PSD: 8.02 dB, right) after an episode of glaucomatocyclitic crisis in the affected eye. [/fig]
|
Ageing transitions in a network of Rulkov neurons
The phenomenon of ageing transitions (AT) in a Erdős-Rényi network of coupled Rulkov neurons is studied with respect to parameters modelling network connectivity, coupling strength and the fractional ratio of inactive neurons in the network. A general mean field coupling is proposed to model the neuronal interactions. A standard order parameter is defined for quantifying the network dynamics. Investigations are undertaken for both the noise free network as well as stochastic networks, where the interneuronal coupling strength is assumed to be superimposed with additive noise. The existence of both smooth and explosive AT are observed in the parameter space for both the noise free and the stochastic networks. The effects of noise on AT are investigated and are found to play a constructive role in mitigating the effects of inactive neurons and reducing the parameter regime in which explosive AT is observed.OPEN
Investigating the dynamical behaviour of an animal brain, formed as an intricate web of large number of interconnected neurons 1 , requires modelling it as a complex dynamical system with underlying network structure. Here, each neuron is a dynamical entity. A complex dynamical system approach enables investigating the time evolution of a large number of interconnected subsystems, the dynamics of each of which are reasonably well understood in isolation but whose collective dynamical behaviour can be significantly different due to the interaction effects. The complexity arises from the interactions, on account of the coupling between the subsystems. Each subsystem is modelled as a node and the mutual interactions due to coupling are modelled as links/edges forming a complex network. One of the most widely investigated complex networked system is the Kuramoto model of phase oscillators. The scale of such complex networked systems can range from 10 to 20 in "low" dimensional oscillator networks 7 to billions of interacting neurons, for example in the human brain model. Phenomena like synchronization, explosive synchronizationand explosive deathare the hallmarks of such complex systems and have been studied across domains ranging from networks of neurons, mechanical oscillatorsto chemical systems.
Investigating the capability of a networked system to operate dynamically even when segments of the system malfunctions is of interest to estimate its robustness to external and/or internal perturbations/malfunctions. Such robustness is observed in many biological systems-such as the brain 1 -which continue to function, maybe sub-optimally, after trauma. In this context, one of the more interesting recent developments has been to study Ageing Transitions (AT) in networks of dynamical systems. In ageing neuronal networks, some nodes (neurons) lose their functionality and become inactive dynamically. When the fraction of inactive nodes is low, the networked system might be able to compensate the loss through coupling and continue to exhibit "active" dynamical behaviour. However, as the fraction of inactive nodes increase, the system collectively transitions to a state where there is no significant "active" dynamical behaviour in the system. AT was first observed in Stuart-Landau (SL) oscillatorsand later in many systems like globally coupled networks, locally coupled oscillators, fractional order SL oscillators, Hindmarsh-Rose neuronsand time-delayed interaction systemsamong others. The nature of AT from the active state of the network to the inactive state can be gradual, monotonic or explosive and depends on several key parameters associated with the problem. A sudden or explosive transition is characterised by a sharp change in the system's dynamical characteristics on a slight change of one of the parameters. These type of transitions are potentially detrimental as this leads to a total loss of capability of the network, without prior warning. There is therefore a need to investigate the parameter regimes at which these explosive transitions occur. An order parameter, which is a measure of the collective dynamical characteristic of a complex network in which the individual components can behave quite differently, is defined to qualitatively and quantitatively investigate the parameter regimes at which these ageing transitions are observed. www.nature.com/scientificreports/ This article focuses on investigating the phenomenon of AT in a network of Rulkov neurons, with the fraction of "inactive" neurons as one of the important parameters. As is well known, neuronal dynamics is typically characterized as a discontinuous system comprising of two distinct states-a silent or inactive state and a spiking or active state. The Rulkov neuron is a simple map based model which captures the discontinuous nature of the neuronal dynamics and unlike differential equation based models, enables numerical studies at significantly cheaper computational costs. This enables carrying out parametric studies for networked systems with a large number of Rulkov neurons. A mean field coupling model is proposed for the neuronal interconnections and the conditions for AT are investigated using measures based on standard order parameters. Subsequently, the dynamics of the networked system is investigated when the mean field coupling model is modified to be stochastic. Parametric studies on the stochastic system reveal that noise could play a constructive role in altering the dynamical characteristics of the network by delaying the onset and the nature of AT. In investigating the dynamics of ageing systems for a given set of parameters, the fraction of "inactive" nodes are taken to be constant as the time-scale of ageing is usually several orders of magnitude larger than the time-scale of the system dynamics.
## Mathematical model
The Rulkov map 29,30 involves two state variables x n and y n , where the former represents the membrane potential of the neuron and is the primary quantity of interest while the latter is an internal variable necessary to capture the dynamics; see Eqs.. The suffix "n" represents the n-th discrete time step. Depending on the choice of the parameters, a solitary neuron exhibits three distinct long-time dynamical states 29 : dead/silent (DS), tonal spiking (TS) and spiking bursting (SB). In this study, the parameters are selected such that the individual neurons exhibit either the DS or TS states. In the case of networked neurons, the map consists of additional terms (see Eqs.that take into account the coupling effects between the i-th and j-th neuron of the network. A symmetric mean field coupling model is adopted where the strength of the coupling is proportional to the mean of the differences between the membrane potential of the i-th neuron and the corresponding membrane potential of all other neurons with which it is directly connected; see Eqs. (6a-6b). This implies that depending on the instantaneous values of the membrane potentials, the nature of the coupling between any two neurons could be either inhibitory or excitatory. The coupling proportionality constant g ij,n is assumed to be time dependent comprising of a constant mean term g m denoting the average coupling strength and an additive fluctuating component, modelled as Gaussian white noise. The fluctuating component accounts for the time dependent modelling uncertainties in the coupling, with the intensity of the fluctuations being represented by D. The network is in the form of a random coupled map lattice (CML), with each node (neuron) having a connection probability P n . Small values of P n lead to a sparsely connected network, while higher values implies a highly connected or dense network. The probability that a particular neuron is inactive (or dead) is defined using parameter P d . For larger values of P d , it is expected that the number of inactive neurons in the network increases, implying an increase in the "age" of the network. The network dynamics is governed by an intricate interplay between the four parameters g m , D, P d and P n . To quantify the collective dynamical characteristics of the network, an order parameter A is defined, which is a normalised measure for the largest amplitude of the oscillations in the steady state condition averaged across all neurons in the network; see Eqs.. A can take values between unity and zero, with zero indicating that none of the neurons are oscillating and hence the network collectively becomes inactive (or dead).
# Results
Exhaustive parametric investigations have been carried out numerically. The representative interesting results that bring out the salient features for both the noise-free and the stochastic cases are discussed next.
Noise free system. The noise free system is when the intensity D = 0 . The collective dynamics is quantified in terms of A as a function of P d for different values of g m and P n . shows the variation of A with respect to P d , for a sparse ER network with P n = 0.1 . www.nature.com/scientificreports/ respectively. The former case essentially considers a set of uncoupled neurons and the latter represents very weak coupling in an already sparse network. In both cases, A is observed to decrease linearly with P d . This is expected as with increasing values of P d , the number of inactive neurons increases, thereby bringing down the value of the averaged normalised order parameter. The zero coupling or low coupling strength implies that the active neurons have little effect on the network. In both cases, A → 0 as P d → 1.
Next, simulations are carried out for an ER network with P n = 0.5 , which represents a more connected network. The variation of A with respect to P d is shown for g m = 0.35 in and for g m = 0.45 in .
These g m values indicate a higher coupling strength. These figures reveal that there is a sudden drop to zero in the value of A at a particular value of P d indicating a sudden transition of the network to the DS.
This sudden discontinuous transition in the state of the network is defined as explosive AT. To gain insights into the collective dynamics of the network, raster plots which show the time histories of the dynamics of each neuron are investigated; see . Here the abscissa is iterate n, the ordinate is the neuron number i and the colour code denotes the values of the state variable x n for each neuron in the network. shows the plots for the case P n = 0.5 , g m = 0.35 and P d = 0.1 for n = 6800-7000. The choice of P d indicates that the regime is to the left of the explosive AT seen in . A homogeneous banded figure pattern is observed, with the bands appearing vertically, indicating that almost all the neurons undergo spiking synchronously. A closer inspection reveals that in between there are neurons which do not undergo any oscillations and these appear as horizontal lines. These are the inactive neurons. The numerical values of x n of these inactive neurons is not zero; in fact, they have different values at the DS which remain constant for all iterates indicating a lack of dynamics. A similar plot for the network having identical parameters but with P d = 0.9-which is in the regime to the right of the explosive AT-is shown in . Here, one can see that the figure comprises of a set of parallel lines indicating that none of the neurons exhibit spiking. All the neurons are therefore in the DS state, and though they have different values of x n , the order parameter A by its definition is zero. It is important to note that as the number of inactive neurons increase in the network, the effects of the active neurons on the network dynamics diminishes www.nature.com/scientificreports/ and beyond a critical threshold value of P d , the higher coupling strength of a reasonably connected ER network ensures that the effects of the active neurons die down with time and the entire network exhibits a DS. Note that shows the time histories only after the effects of the transients have died down. Also, unlike in other systems associated with explosive transitions, hysteresis is not observed here. A parametric study is carried out to investigate in detail the behaviour of the network as P d , g m and P n are varied. -c show the contours of the order parameter A as a function of P d and g m for three different values of P n , corresponding to network topologies that are sparse, intermediate and dense.
The colour-map is same for all cases, with yellow denoting A → 1 and black denoting A → 0 . The black region in these figures denote the parameter regime where the network is globally in the DS. An explosive AT is observed where the colour gradients are sharp. The sharp boundary, in all three cases, indicates the boundary of explosive AT in the parameter space with the colour map indicating the size of the explosive transition. It is interesting to note that the figures look qualitatively similar in all the three cases, indicating that the collective network behaviour is universal irrespective of the topology of the ER network. It is seen that irrespective of the ER network topology or the coupling strength, the critical threshold value for P d required to observe AT is approximately 0.4. This is shown by the horizontal dotted lines in these figures. The vertical dotted lines demarcate the parameter regime such that the region left of this line exhibits smooth transition and to the right shows explosive AT. It is therefore seen that g m should be at least 0.2 for explosive AT to occur, irrespective of the ER network topology.
The critical "death" or "inactivation" probability for P d , denoted by p c , beyond which the network dynamics explosively transitions to a global DS can be quantitatively estimated through the gradient measure γ = �A/�P d . As is obvious, γ shows a high value when there is a discontinuous jump in A as P d is changed. shows the variation of γ with P d for the case P n = 0.5 , g m = 0.5.
From the spike, p c is identified to be 0.5 and this can be corroborated from the inset in . It is interesting to note that the values of γ left of the spike are small but exhibit irregular fluctuations, while the values are identically equal to zero to the right of the spike. This indicates that for P d > p c , the network has globally attained the DS. The variation of p c with g m for the same topology of the ER network is shown in . The sudden It is also observed that p c does not take values less than 0.4 confirming the earlier observation that explosive AT is not observed for P d < 0.4 . The monotonically decreasing nature of the curve underlines the fact that with an increase in the coupling strength and in turn a stronger coupling of the constituents in the network, the global DS of the network is achieved with lower P d . This seems to imply that the DS is a stronger attractor than the TS. shows the variation of p c with respect to both g m and P n confirming that these observations are irrespective of the topology of the ER network.
Stochastic system. Stochasticity is introduced in the system through additive noise to the mean coupling strength g m by considering D = 0 in Eq.. The noise is modelled as a discretized Gaussian white noise process such that the additive noise at each iterate ζ n is standard normal and E[ζ n ζ m ] = δ nm where δ nm is the Kronecker delta. Here, E[ · ] is the expectation operator. As E[g ij,n ] = g m , the noise does not affect the average coupling strength in the ER network. Though the noise is additive with respect to g m , it enters the system as multiplicative noise as g ij,n is a system parameter. presents the variation of A as a function of P d , when P n = 0.5 for the stochastic network for (a) g m = 0.1 and (b) g m = 0.85.
The corresponding variation for the noise-free network ( D = 0 ) is also shown in the same figures for comparison. A smooth transition is observed in , where A → 0 as P d → 1 . An important observation is that A for the stochastic network is greater in comparison to the noise free system for almost the entire regime indicating that the amplitude of the oscillations, on an average, increase with noise intensity D. Explosive AT is observed for higher values of the mean coupling strength for both the noise free as well as the stochastic case as can be seen in . The stochastic network however does not exhibit global DS behaviour post explosive AT, unlike the noise free network. As characterised by the measure A, as P d is further increased beyond p c , there is a steady decline in the oscillations and a global DS is attained as P d → 1 . The magnitude of the oscillations post explosive AT are higher for larger noise intensities.
An important observation from is that noise delays the onset of the explosive AT. This is clearly evident from the figure where the explosive AT for D = 0 is observed to be at P d ≈ 0.4 , while as D becomes larger, explosive AT occurs at higher values of P d . To gain further insights into the collective dynamics of the network, the time histories of the membrane potential x i,n for each neuron is investigated through raster plots; see -b. shows the plot for D = 0.05 , P n = 0.5 , g m = 0.85 , P d = 0.1 and for n = 6800-7000 and denotes a regime prior to explosive AT. Note that the raster plots are shown only for a small segment in time after the initial transients have died down. The banded pattern observed is similar to , indicating that even for the stochastic network all the active neurons exhibit synchronous spiking. The horizontal lines are the inactive neurons having a constant value of x i,n for all n; these values are however different for each of the inactive neurons. However, unlike the noise free network, the vertical bands are not equally spaced indicating that the spikings are no longer regular for the neurons exhibiting spiking. A similar raster plot for the stochastic network with identical parameters but for P d = 0.6 , is shown in . As observed in , this corresponds to post explosive AT regime but the network does not attain global DS. A close inspection of www.nature.com/scientificreports/ explosive AT regime are observed to be inherently noisy whose amplitude as well as the mean value decrease with an increase in P d . A parametric study is next carried out to investigate the behaviour of the stochastic network as P d , g m and D are varied. shows the contour map for A in the g m − P d parameter space for three different cases of noise intensities. The colour-map scale is same for all cases, with yellow denoting A → 1 and black denoting A → 0 . For D = 0.001 , a sharp boundary for the black region is observed in indicating explosive AT and that A is (or close to) zero in the regime post explosive AT. This figure is similar to ; the bounds on g m and P d for which the sharp boundary is encountered are identical. The contour maps shown in ,c, for D = 0.05 andD = 0.07 respectively, are however qualitatively different. These figures still show a sharp boundary in the g m − P d space, indicating explosive AT, but post explosive AT, the order parameter A does not go to zero. An inspection of reveals that no explosive AT is observed for P d < 0.4 for the entire parameter regime. However, as D increases, the minimum values for g m at which explosive AT is observed, increases. A comparison of ,c reveal that the regime in the range of 0.2 < g m < 0.4 which showed explosive AT for low values of D, exhibit smooth AT as D increases. This implies that multiplicative noise can not only increase the region in the g m − P d parameter space at which dynamic neuronal activities are observed, explosive transitions which result in sudden loss of the functionality of the system is also impeded.
These observations are true for all network connectivities P n as has been shown in , where the variation of critical inactivation probability p c is shown in the g m − P n parameter space for D = 0.01 and D = 0.07 . These two cases represents intermediate and high noise intensity.
An additional case that has been studied is to investigate the effects of inhomogeneity associated with each neuron. This was carried out by considering the parameter σ i , which has been so far assumed to be identical for all neurons, to have small variations about the mean value σ . Mathematically, this is modelled by adding a small random perturbation χξ i to the ith neuron, where ξ i is sampled from a standard normal distribution and χ is the strength of the perturbation. Simulations were carried out for χ = 0.01 , and as σ is taken to be 0.6, the variability in σ i ranged from 0.57 to 0.63 with 99.99% probability. Numerical simulations revealed no appreciable www.nature.com/scientificreports/ qualitative or quantitative differences in the results. This can be explained by the fact that small changes in σ are not strong enough to enable a transition of the Rulkov neuron from DS to TS or vice versa. Considering higher values of χ that enable such transitions would imply that P d can no longer be treated as an independent parameter in the model and one would need to instead consider χ as the independent parameter. Instead, considering P d as the independent parameter enables taking into account implicitly the larger variabilities in the parameters associated with a neuron.
# Discussion
The phenomenon of explosive AT in an ER network of coupled Rulkov neurons has been studied, with network connectivity, coupling strength and the fractional ratio of inactive neurons being the parameters of interest. This has been investigated for both the noise free network as well as the stochastic network. In the latter case, the mean inter-neuronal coupling strength have been superimposed with Gaussian white noise at each iterate. An order parameter has been defined for quantifying the dynamical state of the network. It has been shown that explosive AT occurs only when at least 40% of the neurons become inactive both for the noise free and the noisy network, and also depends on the mean coupling strength. For the noisy network, it has been shown the regime at which explosive AT occurs in the parameter space reduces with increasing noise intensity. Moreover, increasing the noise intensity prevents the network from going to DS. However, while noise appears to play a constructive role in preventing explosive AT in certain regimes, the neuronal dynamics in regimes post explosive AT even though do not attain DS, they are noisy and irregular unlike the regular spiking nature expected of neurons. The observations from this study can pave the way for more detailed analysis for probable treatments of impaired neuronal networks.
# Methods
Single neuron. , involving two state variables x n and y n , is given by
Here, x n represents the membrane potential of the neuron and is the primary quantity of interest. The variable y n does not have any particular biological connotation but is important to capture the neuronal dynamics and therefore can be thought of as an "internal" variable. The suffix "n" represents the n-th discrete time step. f α (x, y) is a nonlinear function, given by α and µ are time-independent parameters that defines the qualitative nature of the dynamics of an isolated neuron, while β n and σ n are time dependent parameters that describe the external influences (like injected current) to the system. The spiking frequency increases with σ n 29 . For a neuron in isolation, these parameters can be treated as constants and can be simplified by putting β n = 0 and σ n = σ ∀ n . Depending on the choice of the parameters, a solitary neuron exhibits three distinct long-time dynamical states(1)
x n+1 = f α (x n , y n + β n ),
(2) y n+1 = y n − µ(x n + 1) + µσ n .
( www.nature.com/scientificreports/ In the SB state, the neuron shows bursts of spikes, which may be chaotic depending on the parameter values. The parameter regime considered in this study is limited to the DS and TS states and can be seen in; here, the black region is the DS state while the yellow region denotes the TS state. Beyond α = 4 , there exists the SB regime which is not depicted here because it is not relevant to this study. where for all i, j = 1, . . . , N . Here, the suffices i, j denote the neuron number, N denotes the total number of neurons in the network, the parameters α i and σ i define the individual dynamical nature of the i th neuron, is the degree of the ith node indicating the number of connections of a particular neuron, G ij is an element of the N × N adjacency matrix of the underlying network, g ij,n is the time dependent coupling strength between the i-th and the j-the neuron at time instant n and σ e and β e denote the coupling constants. When σ e = 1 and β e = 1 (which implies that σ i,n = β i,n ∀ i, n ), the model reduces to the original coupling model discussed by for N = 2 . This also implies a symmetric coupling with respect to both x i,n and y i,n . This is a mean-field type coupling which is evident from the fact that the coupling term is a mean of the differences between the membrane potential x i,n of the i-th neuron and the corresponding membrane potentials x j,n of all other neurons it is directly connected to, weighted with a coupling strength g ij,n . Non-symmetric couplings can also be generated by choosing different values for σ e and β e .
[formula] 3) f α (x, y) = α (1−x) + y, x ≤ 0 α + y, 0< x < α + y −1, x ≥ α + y,(a) [/formula]
## Network of neurons.
The most general form of the time-dependent coupling strength g ij,n is given by where g m is the global mean coupling strength, ζ n represents the discretized white noise process at iterate n and D is the noise intensity. For the noise-free case, D = 0 , which implies www.nature.com/scientificreports/ Order parameter. For a network with given values of g m , D, P n -the connection probability of the network-and P d -the probability of a particular neuron being inactive, the collective dynamical state is characterised in terms of an order parameter measure defined as where, for a given value of P d , defines a measure of the largest amplitude of oscillation averaged across all neurons for specified values of P d , g m , P n and D. Here, max n (x i,n ) and min n (x i,n ) indicates the maximum and minimum of x i -the potential corresponding to the ith neuron over all time steps n. The denominator in Eq. (10) refers to the maximum value of a(P d ) when P d is varied from 0 to 1, while P n , g m and D are kept constant. The measure A is therefore a normalized average amplitude measure. These order parameters are similar to those studied for similar systems in literature.
[formula] (4) x i,n+1 = f α i (x i,n , y i,n + β i,n ),(5)y i,n+1 = y i,n − µ(x i,n + 1) + µ(σ i + σ i,n ), (6a) σ i,n = 1 k i j� =i g ij,n σ e (x j,n − x i,n ) if k i � = 0 0 if k i = 0 (6b) β i,n = 1 k i j� =i g ij,n β e (x j,n − x i,n ) if k i � = 0 0 if k i = 0 (7) k i = N j=1 G ij (8) g ij,n = g m + Dζ n G ij ,(9) [/formula]
Simulation details. The network structure is taken to be the Erdős-Rényi (ER) random graph with N = 2 × 10 3 nodes. The connection probability P n is varied from 0.1 to 0.9, thereby covering the whole range of sparse and dense random networks. The initial conditions for each node are assumed to be random and sampled from a uniform distribution between [−1, 1] . The map is iterated over 8 × 10 3 iterates. In computing A, data from the first 5 × 10 3 iterates are discarded and only the last 3 × 10 3 are used. This is carried out to ensure that the transient effects have died down and only the values after stationarity has been achieved are used in computing the quantitate measures used for the analysis. In all the simulations, µ is set to 0.001 to enable spiking state in the neurons. For the parametric analysis in the g m − P d parameter space, a grid is defined by varying g m between [0, 1] in steps of 0.05 and P d is varied between [0, 1] in steps of 0.01. Simulations are carried out for each of these grid points. The noise Dζ n was generated directly using the gsl_ran_gaussian() function from the GNU Scientific Library (GSL), which is an open-source and freely available library. The parametric studies were conducted using a direct grid search type algorithm within the predefined parameter range. All the simulations were performed in double precision for accuracy using C++ programs which were further parallelized using OpenMP for improved performance. All the plots were generated using the plotting capabilities of either MAT-LAB R2020b or gnuplot 5.2.
## (10)
A ≡ A(P d ; g m , P n , D) = a(P d ) max
[formula] 0<P d <1 [a(P d )] ,(11)a(P d ) = 1 N N i=1 max n (x i,n ) − min n (x i,n ) [/formula]
|
Suboptimal endoscopic cancer recognition in colorectal lesions in a national bowel screening programme
## Message
The worldwide implementation of bowel cancer screening programmes (BCSPs) results in a growing number of early T1 colorectal cancers (T1 CRCs). Successful treatment of T1 CRCs starts with accurately recognising these lesions during endoscopy. This study performed in the Dutch BCSP showed that endoscopists correctly diagnosed T1 CRCs in only 39% of 92 cases (95% CI 30 to 49) and that this limited diagnostic accuracy of optical diagnosis resulted in different treatment outcomes. In patients with T1 CRCs that were optically not diagnosed as cancer and treated locally, adjuvant surgery was performed in 41% of cases, compared with 11% of patients with T1 CRCs that were correctly optically diagnosed (p=0.02).
## In more detail
In this prospective multicentre study (trial registration number: NTR4635, NCT02407925), endoscopists accredited for fecal immunochemical test (FIT)-positive colonoscopies within the Dutch BCSP were trained in optical diagnosis with our validated National Institute for Health and Care Excellence (NICE)-WASP (Workgroup serrAted polypS and Polyposis) module (online supplementary material 1 for full methods). 1 2 A total of 27 endoscopists completed the training successfully and entered the prospective study, in which the endoscopists as well as pathologists reported their findings in a predefined structure. This facilitated high-quality data collection and ensured collection of exact data on all aspects of each detected lesion as location, size, Paris morphology, optical diagnosis (colorectal cancer (CRC), adenoma, hyperplastic polyp, sessile serrated lesion or other), endoscopic treatment method (biopsy, cold or hot snare polypectomy, endoscopic mucosal resection, biopsy for diagnosis or no treatment), completeness of resection and whether a tattoo was placed. Participating endoscopists recorded optical diagnosis using narrow band imaging in all consecutive FITpositive colonoscopies for the Dutch BCSP during 1 year. Local treatment was defined as endoscopic resection or transanal endoscopic microsurgery. For all patients initially diagnosed with T1 CRCs, the original H&E staining slides were collected from local hospitals and reviewed. A specialist GI pathologist performed a blinded pathology review on the original slides to corroborate initial diagnosis and provide complete histological information on all T1 CRCs. [bib_ref] Prognostic factors in colorectal carcinomas arising in adenomas: implications for lesions removed..., Haggitt [/bib_ref] [bib_ref] Management of early invasive colorectal cancer. risk of recurrence and clinical guidelines, Kikuchi [/bib_ref] [bib_ref] Colorectal carcinoma: diagnostic, prognostic, and molecular features, Compton [/bib_ref] [bib_ref] Recommendations for reporting tumor budding in colorectal cancer based on the International..., Lugli [/bib_ref] Furthermore, for patients with suspected or established CRC, additional information on follow-up colonoscopies, adjuvant local or surgical therapy, and histopathological outcomes of adjuvant therapy was collected until January 2019.
Between February 2015 and February 2017, 3622 colonoscopies for the Dutch BCSP were performed. Optical diagnosis and histopathology outcome were available for 10 004 lesions (figure 1). In total, 274 patients were diagnosed with CRC, including 90 patients with 92 T1 cancers (online supplementary table 1). The sensitivity for optical diagnosis of CRC was 79.0% (95% CI 73.7% to 83.6%), while the negative predictive value was 99.4% (95% CI 99.3% to 99.5%). Diagnostic test accuracies stratified for confidence level of optical diagnosis can be found in online supplementary table 2. Of 92 T1 cancers, 36 (39.1%, 95% CI 29.1 to 49.9) were correctly optically diagnosed as cancer. Of 56 T1 CRCs optically not diagnosed as cancer (figure 2), 11 (20%) were resected in a piecemeal fashion. In 38 patients, an additional colonoscopy had to be performed to mark the previous endoscopic resection site of the T1 cancer after histology outcome. Adjuvant oncological surgery after local treatment was performed for 2 of 18 (11%) T1 CRCs correctly optically diagnosed as cancer compared with 22 of 54 (41%) of those that were not (p=0.02). Overall, direct or adjuvant surgical treatment after local treatment was performed in 20 of 36 (56%) correctly optically diagnosed T1 CRCs and 24 of 56 (43%) T1 CRCs that were not optically diagnosed as cancer (p=0.29). More details can be found in online supplementary table 3. In 84 of these 90 patients, a median of 2 (IQR 1-2) follow-up exams were performed during a median endoscopic follow-up of 16 months (IQR 12-24). Three patients refused follow-up colonoscopies and three patients were lost to follow-up. One patient died from metastasised T1 CRC 12 months after initial diagnosis, and another patient with metastasised T1 CRC is still alive. None of the patients had a local recurrence. There was no difference in terms of survival and treatment outcomes between the optically correctly and incorrectly diagnosed T1 CRCs.
## Comments
BCSPs aim to detect CRC at an early stage. In a recent report of the Dutch FIT-based BCSP, 40% of all CRCs were T1. [bib_ref] Stage distribution of screen-detected colorectal cancers in the Netherlands, Toes-Zoutendijk [/bib_ref] Patients with T1 cancers with favourable histological characteristics are at low risk for lymph node metastasis (6%-27%), [bib_ref] Long-Term recurrence-free survival after standard endoscopic resection versus surgical resection of submucosal..., Belderbos [/bib_ref] [bib_ref] Predicting lymph node metastasis in pT1 colorectal cancer: a systematic review of..., Bosch [/bib_ref] [bib_ref] Population-based analyses of lymph node metastases in colorectal cancer, Ricciardi [/bib_ref] and the majority of these lesions can be cured with endoscopic treatment. [bib_ref] Management of malignant colonic polyps: a population-based analysis of colonoscopic polypectomy versus..., Cooper [/bib_ref] [bib_ref] Endoscopic resection of high-risk T1 colorectal carcinoma prior to surgical resection has..., Overwater [/bib_ref] [bib_ref] Oncological outcome after local excision of rectal carcinomas, Borschitz [/bib_ref] [bib_ref] Long-term outcomes after resection for submucosal invasive colorectal cancers, Ikematsu [/bib_ref] Several histological characteristics predict the risk of lymph node metastases. These risk factors can only be assessed when the lesion is resected en bloc. Therefore, successful treatment of these lesions starts with suspecting a T1 cancer, either with superficial submucosal (SMs) or deep submucosal (SMd) invasion.
Our study demonstrates that optical diagnosis of T1 CRC among accredited endoscopists in the Dutch BCSP can be further improved. Overall sensitivity for optical diagnosis of all stages of CRC was 79.0%, while the negative predictive value was 99.4%. This suboptimal sensitivity relied heavily on a limited sensitivity of optical diagnosis of T1 CRCs, which were optically diagnosed correctly as cancer in only 39.1% (95% CI 29.1 to 49.9). The limited diagnostic accuracy also resulted in suboptimal treatment decisions, as 20% of T1 CRCs optically not diagnosed as cancer were removed in a piecemeal fashion, thus prohibiting optimal histological analysis. As a result, adjuvant surgical treatment after local treatment was more frequently indicated and performed for patients with T1 CRCs that were not correctly optically diagnosed (41% vs 11%, p=0.02). However, in terms of survival and treatment outcomes, there were no differences between the two groups, although the median follow-up was limited to 16 months. Moreover, patient characteristics and preferences may have influenced the decision for adjuvant surgical treatment.
In a recently published prospective Spanish study in which 58 endoscopists from community and university hospitals examined over 2000 lesions of at least 10 mm in size using the NICE classification, the NICE classification identified lesions with SMd with 58% sensitivity (95% CI 48 to 69) and negative predictive on March 18, 2020 by guest. Protected by copyright.
http://gut.bmj.com/ value of 98% (95% CI 97 to 99). [bib_ref] Accuracy of the narrow-band imaging international colorectal endoscopic classification system in identification..., Puig [/bib_ref] In another real-time Dutch study with endoscopists trained in optical diagnosis of CRC, which included large non-pedunculated lesions of >20 mm, a much higher sensitivity for optical diagnosis was reported, namely 79% (95% CI 64 to 89). [bib_ref] Multicentre prospective evaluation of realtime optical diagnosis of T1 colorectal cancer in..., Backes [/bib_ref] However, the positive predictive value in this study was quite low at 69% (95% CI 57 to 78), which resulted in unnecessary surgery. This was also shown in other studies. Hence, we can conclude that incorrect optical diagnosis for predicting SM and SMd invasion results in suboptimal use of endoscopic and surgical treatment options, resulting in both overtreatment and undertreatment. On the other hand, there is a narrow gap between T1 cancer detected and diagnosed by endoscopy that is curable by endoscopic treatment methods. In most series, up to 40% of patients with T1 cancers resected by endoscopic mucosal resection, endoscopic submucosal dissection or endoscopic full-thickness resection have to undergo additional surgery anyway due to high-risk histological criteria. [bib_ref] Endoscopic submucosal dissection of malignant non-pedunculated colorectal lesions, Rönnow [/bib_ref] [bib_ref] Endoscopic full-thickness resection for early colorectal cancer, Kuellmer [/bib_ref] [bib_ref] Risk Stratification for Covert Invasive Cancer Among Patients Referred for Colonic Endoscopic..., Burgess [/bib_ref] Observational studies with validated training in optical diagnosis of T1 cancers for community gastroenterologists are needed, aiming to achieve high sensitivity and negative predictive value for endoscopic prediction of T1 cancers. Preferably, endoscopists would also achieve high accuracies for differentiating between SMs and SMd. To determine whether lesions are suitable for endoscopic resection, a structured lesion assessment seems useful. The recent British Society of Gastroenterology and European Society of Gastrointestinal Endoscopy guidelines suggest to use the NICE and/or Kudo classification with highdefinition virtual chromoendoscopy techniques to assess SM invasion, as these have shown good interobserver agreement and are easily adapted in clinical practice. This study has several limitations. We included all detected lesions, irrespective of size and morphology, for calculating diagnostic test accuracies. As the prevalence of CRC within small lesions is low, the negative predictive value of optical diagnosis of CRC is likely overestimated. On the other hand, every colorectal lesion can harbour CRC and should therefore be characterised before applying treatment. Furthermore, we excluded lesions that were referred for endoscopic removal because this was often performed by other specialist endoscopists on dedicated programmes, possibly introducing selection bias . Although optical diagnosis was recorded, participating endoscopists did not state if SMs or SMd was suspected. Besides, endoscopists were not formally trained in endoscopic recognition of the depth of invasion. On the other hand, endoscopists participating in this study were accredited for performing colonoscopies for the Dutch BCSP. [bib_ref] Quality assurance of colonoscopy within the Dutch national colorectal cancer screening program, Bronzwaer [/bib_ref] We believe that the results of our study can therefore be extrapolated to other organised screening programmes in which general gastroenterologists perform colonoscopies.
In this prospective study, approximately two-thirds of the T1 CRCs were not recognised during consecutive FIT-positive colonoscopies for the Dutch BCSP, leading to suboptimal use of endoscopic and surgical treatment options. Training in structured lesion assessment is needed to improve endoscopic recognition of T1 CRCs and ensure optimal treatment strategies.
acknowledgements We thank Joanne Verheij, MD, PhD, for participating in the specialist pathology review. We thank all endoscopy nurses working in participating centres for their invaluable help with data collection.
[fig] Figure 1: STARD flow chart describing study flow. Between February 2015 and February 2017, 28 participating endoscopists performed 3622 colonoscopies for the Dutch BCSP and detected 10 859 lesions during these colonoscopies. The figure shows the flow through the study along with the primary outcome of optical diagnosis of T1 CRC. Reasons for exclusion are noted. *Depicts the number of T1 CRCs of the total group of CRCs. BCSP, bowel cancer screening programme; CRC, colorectal cancer; STARD, Standards for Reporting Diagnostic Accuracy; T1 CRC, T1 colorectal cancer. [/fig]
[fig] Figure 2: White light (A,C) and corresponding narrow band imaging (B,D) pictures of histologically confirmed T1 colorectal cancers optically diagnosed as adenomas.27 [/fig]
|
Sleep Duration, Lifestyle Intervention, and Incidence of Type 2 Diabetes in Impaired Glucose Tolerance
OBJECTIVE -Both short and long sleep duration have frequently been found to be associated with an increased risk for diabetes. The aim of the present exploratory analysis was to examine the association between sleep duration and type 2 diabetes after lifestyle intervention in overweight individuals with impaired glucose tolerance in a 7-year prospective follow-up.RESEARCH DESIGN AND METHODS -A total of 522 individuals (aged 40 -64 years) were randomly allocated either to an intensive diet-exercise counseling group or to a control group. Diabetes incidence during follow-up was calculated according to sleep duration at baseline. Sleep duration was obtained for a 24-h period. Physical activity, dietary intakes, body weight, and immune mediators (C-reactive protein and interleukin-6) were measured.RESULTS-Interaction between sleep duration and treatment group was statistically significant (P ϭ 0.003). In the control group, the adjusted hazard ratios (HRs) (95% CI) for diabetes were 2.29 (1.38 -3.80) and 2.74 (1.67-4.50) in the sleep duration groups 9 -9.5 h and Ն10 h, respectively, compared with for that of the 7-8.5 h group. In contrast, sleep duration did not influence the incidence of diabetes in the intervention group; for sleep duration groups 9 -9.5 h and Ն10 h, the adjusted HRs (95% CI) were 1.10 (0.60 -2.01) and 0.73 (0.34 -1.56), respectively, compared with that in the reference group (7-8.5 h sleep). Lifestyle intervention resulted in similar improvement in body weight, insulin sensitivity, and immune mediator levels regardless of sleep duration.CONCLUSIONS -Long sleep duration is associated with increased type 2 diabetes risk. Lifestyle intervention with the aim of weight reduction, healthy diet, and increased physical activity may ameliorate some of this excess risk.
I t has become increasingly evident that abnormal sleep patterns, such as short and long sleep durations, are associated with increased morbidity and mortality [bib_ref] Mortality associated with sleep duration and insomnia, Kripke [/bib_ref]. The relationship between increased risk for type 2 diabetes and either short or long sleep has frequently been demonstrated in both men and women [bib_ref] A prospective study of selfreported sleep duration and incident diabetes in women, Ayas [/bib_ref] [bib_ref] Sleep duration is associated with an increased risk for the prevalence of..., Tuomilehto [/bib_ref]. The underlying mechanisms are still largely unknown. In an experimental study, sleep deprivation has been shown to have harmful effects on metabolic and endocrine functions [bib_ref] Impact of sleep debt on metabolic and endocrine function, Spiegel [/bib_ref]. However, there are no plausible physiological or psychosocial explanations for the association between long sleep and diabetes. In fact, some studies have observed a stronger association between habitual long sleep and diabetes than that for short sleep [bib_ref] A prospective study of selfreported sleep duration and incident diabetes in women, Ayas [/bib_ref]. There is no evidence or consensus whether the relationship between long sleep and diabetes is causal or simply reflects confounding by other underlying factors that affect sleeping habits. However, there is growing evidence that many of the hypothesized factors for long sleep are strongly associated with obesity and metabolic syndrome [bib_ref] Depressive symptoms and metabolic syndrome: is inflammation the underlying link, Capuron [/bib_ref]. The objective of the present exploratory analysis was to examine the effect of lifestyle intervention on the association between sleep duration and type 2 diabetes in individuals with impaired glucose tolerance (IGT) in the Finnish Diabetes Prevention Study (DPS) [bib_ref] Finnish Diabetes Prevention Study Group. Sustained reduction in the incidence of type..., Lindströ M [/bib_ref].
## Research design and
METHODS -This report is a substudy of a larger lifestyle intervention trial, in which 522 overweight individuals with IGT were randomly allocated to an intensive diet-exercise group (n ϭ 265) or a control group (n ϭ 257). At baseline, participants' mean Ϯ SD BMI was 31.1 Ϯ 4.5kg/m 2 and age was 55 Ϯ 7 years. There were no significant differences in clinical features between the two groups. More details on the study design can be found in previously published reports [bib_ref] Finnish Diabetes Prevention Study Group. Sustained reduction in the incidence of type..., Lindströ M [/bib_ref].
## Lifestyle intervention
The participants in the intensive intervention group were given detailed, individualized counseling to achieve the lifestyle goals: 1) weight reduction Ն5%, 2) Ͻ30% of the daily energy intake from fat, 3) Ͻ10% of the daily energy intake from saturated fat, 4) fiber intake Ն15 g/1,000 kcal, and 5) moderately intense physical activity 30 min daily. The participants in the control group were given general oral and written advice about diet and exercise at baseline, but no specific individualized advice was offered to them. The intervention period lasted for a median of 4 years, after which a postintervention follow-up lasting for a median of 3 years was carried out [bib_ref] Finnish Diabetes Prevention Study Group. Sustained reduction in the incidence of type..., Lindströ M [/bib_ref].
## Sleep duration and anthropometric assessments
Before each annual examination, participants completed an activity diary covering all activities (divided into eight categories) during the past 24 h. Duration of activities was recorded to the nearest half hour. Time spent sleeping or resting in bed was considered as an individual's daily sleep duration. Measurements of height, weight, waist circumference, and blood pressure have been described in detail previously [bib_ref] Finnish Diabetes Prevention Study Group. Sustained reduction in the incidence of type..., Lindströ M [/bib_ref].
## Assessment of dietary intake and physical activity
Study participants completed 3-day food records at the baseline and before each annual visit. The physical activity was calculated based on the Leisure-Time Physical Activity questionnaire [bib_ref] Intra-person variability of various physical activity assessments in the Kuopio Ischaemic Heart..., Lakka [/bib_ref]. The details of both dietary intake and physical activity assessments and calculations can be found in earlier publications [bib_ref] Finnish Diabetes Prevention Study Group. Sustained reduction in the incidence of type..., Lindströ M [/bib_ref].
## Blood sampling and laboratory measurements
Blood was taken from the antecubital vein with the participant in a sitting position and allowed to clot at room temperature for 30 -60 min. After centrifugation at 8,000 -11,000g, sera were stored at Ϫ70°C for analyses although for logistic reasons, storage at Ϫ20°C was allowed for a maximum of 3 months. The annually measured biochemical parameters included an 75-g oral glucose tolerance test after 12 h of fasting in the morning with fasting and 2-h plasma glucose measurements. Diabetes was defined according to the World Health Organization 1985 criteria. The diagnosis of diabetes had to be confirmed by a second oral glucose tolerance test. Serum concentrations of high-sensitivity C-reactive protein (CRP) were assessed by an immunonephelometric assay (Dade Behring, Marburg, Germany) [bib_ref] for the Finnish Diabetes Prevention Study Group. Systemic immune mediators and lifestyle..., Herder [/bib_ref]. Interleukin-6 (IL-6) concentrations in serum were determined by enzyme-linked immunosorbent assay using recombinant IL-6 and an antibody pair (Sanquin, Amsterdam, Netherlands) [bib_ref] for the Finnish Diabetes Prevention Study Group. Systemic immune mediators and lifestyle..., Herder [/bib_ref].
# Statistical methods
Sleep duration was grouped as Յ6.5 h, 7-8.5 h, 9 -9.5 h, and Ն10 h. Diabetes incidence per 100 person-years was calculated in each sleep duration group. The Cox proportional hazards model was used to compare the development of diabetes in different sleep duration categories. The average sleep duration for an adult in Finland is 7.5 h (13), and therefore the group 7-8.5 h was used as reference category. In a multivariate Cox model, the relationship between diabetes and sleep duration was further explored by taking into account possible confounding factors including 1) age, sex, study center, and BMI at baseline and 2) age, sex, BMI, study center, smoking, alcohol intake, hypertension medication, leisure-time physical activity at baseline, and 1-year change in body weight. The interaction between treatment group and sleep duration at baseline was tested to assess whether the intervention modified the relationship between sleep duration and diabetes incidence. In addition, the effect of treatment and sleep on diabetes was analyzed using the annual sleep duration assessments with the Cox proportional hazards regression model, treating the repeated measurements on sleep duration as a time-varying covariate. Analyses were done with the statistics package Stata (release 9.2, 2005; StataCorp, College Station, TX).
# Results
-Age was directly associated with sleep duration in both treatment groups [fig_ref] Table 1 -: Characteristics of the DPS study participants by sleep duration at baselineData are... [/fig_ref]. The patients with longer sleep duration were more likely to have a nap during the daytime and medication for hypertension; otherwise the treatment groups had no statistically significant differences in baseline characteristics between the different sleep duration groups. However, there was a tendency toward more frequent smoking, alcohol usage, and physical activity in long sleepers (Ն10 h) compared with other sleep duration groups.
Sleep duration at baseline was not significantly associated with body weight [fig_ref] Table 1 -: Characteristics of the DPS study participants by sleep duration at baselineData are... [/fig_ref] and did not predict weight change over 3 years [fig_ref] Table 2 -: Changes over 1-year follow-up among the DPS participants by sleep duration at... [/fig_ref]. Neither were there significant differences in CRP or IL-6 levels at the follow-up visit between different sleep duration groups. The intervention resulted in parallel improvements in body weight, waist circumference, immune mediator levels, and insulin sensitivity over the follow-up in all of the sleep groups.
Small changes in the mean sleeping time during the follow-up were observed [fig_ref] Table 2 -: Changes over 1-year follow-up among the DPS participants by sleep duration at... [/fig_ref]. The long sleepers (Ն10 h) in the control group slept even more, whereas the sleeping time was reduced in long sleepers in the intervention group. However, the changes were minor, and there was no significant shifting from one sleep duration group to another in either the control or intervention group. Over the 1-year follow-up, the proportion of long-sleeping participants (Ն10 h) taking a nap during the daytime decreased from 74% to a mere 30% with similar changes in both treatment groups. During the follow-up, the physical activity of long sleepers was significantly reduced in the control group only, but the interaction term indicates that the difference was nonsignificant between both treatment groups. The changes in diet based on food diaries in the intervention group were similar in different sleep duration groups.
A total of 182 participants developed diabetes during the follow-up [fig_ref] Table 3 -: Incidence rates for type 2 diabetes by sleep duration at baselineData in... [/fig_ref]. The interaction between treatment group and sleep duration at baseline was statistically significant (adjusted P ϭ 0.003). In the control group only, there was a statistically significant difference in diabetes incidence rates between the sleep duration groups (log-rank test P ϭ 0.001) [fig_ref] Figure 1 -: The cumulative incidence of type 2 diabetes in follow-up time by sleep... [/fig_ref]. For the sleep duration groups 9 -9.5 and Ն10 h, the adjusted hazard ratios (HRs) were 2.29 (95% CI 1.38 -3.80) and 2.74 (1.67-4.50), respectively, compared with that in the reference group (7-8.5 h sleep) [fig_ref] Table 3 -: Incidence rates for type 2 diabetes by sleep duration at baselineData in... [/fig_ref]. In contrast, no significant differences in diabetes incidence among the sleep duration groups in the intervention group were seen. In the intervention group, the adjusted HRs were 1.10 (0.60 -2.01) and 0.73 (0.34 -1.56) for the sleep duration groups 9 -9.5 and Ն10 h, respectively, compared with that in the reference group (7-8.5 h sleep) [fig_ref] Table 3 -: Incidence rates for type 2 diabetes by sleep duration at baselineData in... [/fig_ref]. A tendency toward an increased diabetes risk in individuals with short sleep (Յ6.5 h) was also observed [fig_ref] Table 3 -: Incidence rates for type 2 diabetes by sleep duration at baselineData in... [/fig_ref]. In a Cox model with the annually repeated sleep duration assessments included as timevarying covariates, the interaction term between treatment group and sleep duration was still statistically significant (adjusted P ϭ 0.028).
Daytime sleepiness, as measured by sleeping and resting in bed during the daytime (0900 -1900), predicted diabetes incidence in the control group but not in the intervention group (P interaction ϭ 0.001). However, when total sleep duration was simultaneously accounted for, daytime sleepiness was no longer associated with diabetes incidence in the control group (P ϭ 0.526).
CONCLUSIONS -In this post hoc analysis of the DPS, we detected a significant association between long sleep duration and incident type 2 diabetes in overweight individuals with IGT in a 7-year follow-up. However, the association was not observed in participants receiving individualized lifestyle intervention. The facts that 1) study participants with different sleep duration at baseline did not differ in anthropometric, metabolic, and immunological parameters and that 2) changes in these parameters were similar in both intervention and control groups suggest that the association between sleep duration and diabetes risk was independent of these factors. Short sleep duration as a possible cause of diabetes incidence is plausible given results from experimental studies in healthy volunteers that have shown sleep restriction to result in glucose intolerance [bib_ref] Impact of sleep debt on metabolic and endocrine function, Spiegel [/bib_ref] , which is a risk factor for the development of insulin resistance. In the present study, there was a tendency toward increased diabetes risk in individuals with short sleep, but it did not reach statistical significance compared with the reference group. This could be because there was a small number of participants with short sleep and thus inadequate statistical power to detect a difference or because the short sleep category was defined somewhat broadly (Յ6.5 h/night) and was not particularly short compared with the typically recommended sleep duration of 7-9/night. Thus far, there is no compelling physiological evidence explaining the association between long sleeping time and increased morbidity and mortality. The link has been observed in epidemiological studies, based on questionnaires without objective sleep measurements. Therefore, whether long sleep represents an increase in actual sleeping time or just increased time in bed cannot be definitely determined. The main underlying causes proposed to potentially explain the relationship have been obstructive sleep apnea (OSA), depression, fatigue, low socioeconomic status, heart diseases, and inactivity [bib_ref] Self-reported sleep duration in Finnish general population, Kronholm [/bib_ref] [bib_ref] Who are long sleepers? Towards an understanding of the mortality relationship, Grandner [/bib_ref] [bib_ref] Correlates of long sleep duration, Patel [/bib_ref]. Of these proposed factors, sleep-disordered breathing including OSA has been found to be independently associated with insulin resistance [bib_ref] Obstructive sleep apnea is independently associated with insulin resistance, Ip [/bib_ref]. In overweight patients with OSA, lifestyle intervention with weight reduction has been found to be an effective treatment [bib_ref] Weight reduction is a curative treatment for patients with mild obstructive sleep..., Tuomilehto [/bib_ref]. An earlier analysis of the DPS revealed that weight reduction also led to a significant improvement in insulin resistance in patients with IGT [bib_ref] Longterm improvement in insulin sensitivity by changing lifestyles of people with impaired..., Uusitupa [/bib_ref]. Altogether, characteristics of the long sleepers have been found to differ from those of short or normal sleepers, and these findings tend to support many of the proposed hypotheses [bib_ref] Self-reported sleep duration in Finnish general population, Kronholm [/bib_ref] [bib_ref] Who are long sleepers? Towards an understanding of the mortality relationship, Grandner [/bib_ref]. The long sleepers in our study were observed to be older and to take naps more frequently and were more likely to be receiving antihypertensive drug therapy than other participants. Antihypertensive drugs, particularly -blockers, have been reported to have central nervous system ef- fects, resulting in tiredness, fatigue, and sleep disorders in some individuals [bib_ref] Drugs that disturb sleep and wakefulness, Schweitzer [/bib_ref].
In the present study, participants with long sleep were more likely to take naps during the daytime; a possible manifestation of daytime sleepiness known to be related with OSA. Over the follow-up period, less long-sleeping participants were observed to take a nap during the daytime; however, the change was equal between the treatment groups. In the intervention group, participants with longsleep at baseline tended to sleep less over follow-up and participants with short sleep at baseline tended to sleep longer, whereas the opposite occurred in the control group. However, the overall changes in sleep durations over the follow-up were relatively small. The interaction term between sleep duration and treatment was still statistically significant, indicating that the effect of intervention was different among the different sleep duration categories.
Many of the factors documented to be associated with long sleep duration are also associated with obesity and the metabolic syndrome [bib_ref] Depressive symptoms and metabolic syndrome: is inflammation the underlying link, Capuron [/bib_ref]. The association between long sleep duration and diabetes has been hypothesized to be due to the effects of proinflammatory cytokines, which have also been found to be elevated in sleep-disordered breathing, depression, and excessive sleepiness, the suggested confounding factors causing long sleep [bib_ref] Elevation of plasma cytokines in disorders of excessive daytime sleepiness: role of..., Vgontzas [/bib_ref] [bib_ref] Sleep duration and snoring in relation to biomarkers of cardiovascular disease risk..., Williams [/bib_ref]. Furthermore, obesity it-self is regarded as a chronic inflammatory condition. The elevated levels of proinflammatory cytokines have been considered to contribute to the increased sleep duration [bib_ref] Sleep duration and snoring in relation to biomarkers of cardiovascular disease risk..., Williams [/bib_ref]. A previous report from the Finnish DPS demonstrated that systemic concentrations of immune mediators were associated with the progression of IGT to type 2 diabetes, and the prevention of type 2 diabetes by lifestyle changes was related to the decline in proinflammatory factors. This decline, on the other hand, was associated with increases in fiber intake and moderate to vigorous leisure time physical activity [bib_ref] for the Finnish Diabetes Prevention Study Group. Systemic immune mediators and lifestyle..., Herder [/bib_ref] [bib_ref] the Finnish Diabetes Prevention Study Group. Anti-inflammatory effect of lifestyle changes in..., Herder [/bib_ref]. We did not observe differences in CRP or IL-6 concentrations between the long sleep duration and the reference group at the baseline. The level of CRP was higher in both short and long sleepers compared with that in the reference group; however, the difference was not statistically signifi-cant. Neither were there any significant changes in CRP or IL-6 during the follow-up in the long sleep duration group compared with the reference group. In the DPS, all of the individuals belonged to a high-risk group with overweight and IGT and therefore do not represent a normal population cohort. Physical activity and dietary intake patterns were similar at the baseline among the different sleep duration strata, and the changes during the follow-up did not either differ among the strata groups. The assessment of sleep duration in our study was based on a self-reported activity questionnaire and not confirmed by objective measurements. Nevertheless, questionnaires have been found to be useful in assessing sleep duration [bib_ref] Comparison between subjective and actigraphic measurement of sleep and sleep rhythms, Lockley [/bib_ref]. The activity diary used in the present study covers all activities divided into eight categories, with one of those categories being sleeping or resting in bed [bib_ref] Intra-person variability of various physical activity assessments in the Kuopio Ischaemic Heart..., Lakka [/bib_ref]. Thus, unfortunately, it is not possible to separate them. However, in the present study the data collected on sleep duration was obtained for a 24-h period, including naps, a possible manifestation of daytime sleepiness. We have shown previously that lifestyle intervention was effective in lowering diabetes risk. The present analyses reveal that the effect of intervention was mainly limited to those with longer baseline sleep duration. Because we observed no interaction between sleep duration and group assignment on dietary and physical activity behavior intervention, this risk reduction apparently did not result from better adherence to intervention. In clinical work more attention should be paid to these conditions; along with visual detection of obesity, physicians may more effectively identify patients at high risk for diabetes with questions about the sleep quality and quantity.
Our results are in line with some previous reports suggesting that habitually long sleep may be a surrogate marker of a condition that is associated with an increased risk for type 2 diabetes [bib_ref] Who are long sleepers? Towards an understanding of the mortality relationship, Grandner [/bib_ref]. It is likely that a number of factors exist to explain the statistical association between long sleep duration and diabetes. Whether all the factors correlate to each other is unclear, but most of them are commonly related to fatigue, inactivity, weight gain, and obesity. In summary, the present results suggest that some of the type 2 diabetes risk among overweight persons with IGT who also have long sleep duration may be ameliorated by weight reduction and an increase in physical activity and healthier diet.
[fig] Figure 1 -: The cumulative incidence of type 2 diabetes in follow-up time by sleep duration groups. A: Control group. B: Intervention group. [/fig]
[table] Table 1 -: Characteristics of the DPS study participants by sleep duration at baselineData are means Ϯ SD unless stated otherwise. *Geometric means. HOMA-IR, homeostasis model assessment of insulin resistance. LTPA, leisure-time physical activity. *Geometric means. [/table]
[table] Table 2 -: Changes over 1-year follow-up among the DPS participants by sleep duration at baselineData are mean changes adjusted for baseline levels of the respective variable, except the daytime nap, which is reported as prevalence at 1-year follow-up. HOMA-IR, homeostasis model assessment of insulin resistance; LTPA, leisure-time physical activity. *P value for test of equal change between the sleep duration groups. †P value for test of interaction between sleep duration and treatment group. ‡P Ͻ 0.05 compared with the 7.0 -8.5 h group. [/table]
[table] Table 3 -: Incidence rates for type 2 diabetes by sleep duration at baselineData in parentheses are 95% CI. *P value: log-rank test for equal incidence over sleep duration categories. †Adjusted for age, sex, BMI, and study center. ‡Adjusted for age, sex, BMI, study center, smoking, alcohol intake, hypertension medication, leisure-time physical activity at baseline, and 1-year change in body weight. [/table]
|
Design and characterization of an eight‐element passively fed meander‐dipole array with improved specific absorption rate efficiency for 7 T body imaging
Objective: To evaluate the transmit efficiency and specific absorption rate (SAR) efficiency of a new eight-element passively fed meander-dipole antenna array designed for body MRI at 7 T, and to compare these values with a conventional directly fed meander-dipole array.Methods:The main radiating element of the passively fed dipole is printed on one side of a dielectric substrate and is capacitively coupled to a shorter feeding element (connected to the coaxial cable) printed on the opposite side of the substrate. The transmit (B 1 + ) field and SAR were simulated on a phantom and on a human voxel model for both a passively fed and a directly fed single element. Two eight-channel arrays containing, respectively, directly and passively fed meander dipoles were then simulated, and experimental B 1 + maps and T 2 -weighted spin echo images of the prostate were obtained in vivo for four healthy volunteers.Results: In simulations, the mean transmit efficiency (B 1 + per square root input power) value in the prostate was~12.5% lower, and the maximum 10 g average SAR was 44% lower for the array containing passively fed dipoles, resulting iñ 15% higher SAR efficiency for the passively fed array. In vivo RF-shimmed turbo spin echo images were acquired from both arrays, and showed image SNRs within 5% of one another.Conclusion: A passive-feeding network for meander-dipole antennas has been shown to be a simple method to increase the SAR efficiency of a multi-element array used for body imaging at high fields. We hypothesize that the main reason for the increase in SAR efficiency is the storage of the strong conservative electric field in the dielectric between the feeding element and the radiating element of the dipole.The passive-feeding approach can be generalized to other dipole geometries and configurations.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
# | introduction
Human MRI at high field strengths is challenging due to significant inhomogeneities in the transmit (B 1 + ) fields that occur as a result of the relatively short RF wavelength in tissue compared with the dimensions of the human body. This is particularly evident in body applications at field strengths of 7 T and higher,which has led to the use of transmit arrays consisting of multiple (typically 8-32 elements) transmit antennas.This approach uses individually controlled transmit channel phases and/or magnitudes. A number of different resonators have been used for the individual elements of these arrays, including loops and microstrips,but in the past few years various forms of dipole antennas have become the predominant geometry. The intrinsic far-field nature of a dipole is advantageous in terms of transmitting electromagnetic energy into deep-lying organs within the body.Dipoles have been shown both theoreticallyand experimentallyto be efficient transmit elements for body imaging at 7 T in particular, and have also recently been used at 10.5 T.After the initial demonstration of the performance of a conventional straight dipole, a substantial amount of research has concentrated on optimizing the dipole geometry in order to improve a combination of the transmit efficiency (B 1 + field per square root input power) and the specific absorption rate (SAR) efficiency (B 1 + field to the square root maximum specific absorption rate [SAR 10g, max ]). For example, the fractionated dipole antenna presented inis a straight printed dipole with several meandered structures included in each leg of the dipole, which act as inductances to increase the electrical length of the dipole. This design produces the lowest SAR 10g, max among current dipoles used for high field imaging. Dipoles can be used as transmit-only or transceiver elements, either alone or in combination with loop antennas as additional receive elements .Fractionated dipoles can be combined in pairs to give independent excitation of the even and odd modes of the dipole pair.For all of the dipole geometries described in the literature, including the meander dipole, the element is directly fed via a partially or fully balanced lumped matching element network consisting of capacitors and/or inductors. However, it is also possible to feed dipole elements passively, ie via inductive and capacitive coupling from a passive structure that is coupled to the dipole.In this study, we evaluate experimentally and in simulations the passive-feeding approach for a 300 mm long meander-dipole antenna used as an array element for prostate imaging at 7 T. The main radiating antenna element was printed on one side of a dielectric substrate, coupled to a shorter passive feeding element printed on the opposite side. Numerical and experimental comparisons were performed with a conventional array containing directly fed meander-dipole antennas.
## | materials and methods
## | dipole fabrication
The passively and directly fed meander-dipole geometries are shown in. The length and geometry of the short feeding dipole (in the passively fed meander-dipole design) was chosen via a parametric study (data not shown), in which the length of the feeding dipole was varied from 90 to 150 mm, in steps of 10 mm, and both a straight and meandered geometry were studied, in order to obtain the maximum SAR efficiency. The design of the directly fed meander dipole is identical to that described previously.The dipoles were each 300 mm long and were fabricated on an FR-4 printed circuit board (PCB) substrate (ɛ r = 4.3, tan δ = 0.025, substrate thickness = 1.5 mm). For a passively fed dipole, one side of the PCB ("back" side) contained a short feeding element (length = 110 mm) with a small meander at the end of each leg. The other side of the PCB ("front" side) contained a larger radiating dipole and was oriented towards the phantom/subject. The passively fed meander dipole had a 33 nH inductor connected between the legs of the short feeding element and two identical 3.9 pF matching capacitors. The directly fed meander dipolehad two identical matching capacitors with a value of 22 pF. For the transmit/receive array eight directly fed and eight passively fed meander antennas with the same total dimensions as described above were constructed. A 20 mm thick polymethylmethacrylate (PMMA) spacer was attached to the face of the dipole.
S 11 was measured for both the directly and passively fed meander dipoles as a function of the distance between the single antenna and the phantom: this distance was varied from 20 to 70 mm in steps of 10 mm. The reference point for both antennas was the feed point, ie the short feeding dipole was at the same distance from the phantom as the feed point of the directly fed dipole. Simulated RF shimming was performed by adjusting the phases of the individual channels to produce the maximum constructive interference of all channels in the prostate region. The phase of the B 1 + field produced by each channel in the prostate was estimated, and the corresponding negative phase applied to the transmit channel was such that the phases of all channels were zero in the prostate, and that the B 1 + fields added constructively. x 380 x 16 mm, voxel size = 0.8 x 0.8 mm, slice thickness = 3 mm, one signal average. A noise correlation matrix was generated by acquiring data with the RF and gradient channels turned off, with the noise correlation coefficient (ρ ij ) defined as
## | electromagnetic simulations and antenna measurements
## | mri measurements
[formula] ρ ij ¼ cov n i n j À Á σ ni σ nj [/formula]
where n i is the noise measured in coil element i. SNR maps were acquired using the method described previously .3 | RESULTSshows simulated and measured B 1 + and SAR 10g, max values, as well as the thermometry measurements from single dipole elements placed on the phantom. For the passively fed meander-dipole antenna, the simulated and measured B 1 + field (normalized to 1 W of accepted power) of a single element is lower than that of the directly fed antenna, while the SAR 10g, max is reduced by 37%and 3F). shows the results of the MRI-based thermal measurements performed for the same input power and the same heating time. The passively fed antenna produced~50% less heating (measured at the point of maximum heating, which was directly under the central point of the dipole) than the directly fed antenna. The maximum temperature increase produced by the directly fed meander antenna was 12°C while the maximum temperature increase for the passively fed meander antenna was 6°C.shows a graph of the B 1 + per square root of unit power (transmit efficiency) and B 1 + per square root of SAR 10g, max (SAR efficiency)
for both directly and passively fed dipoles plotted along the central line in the phantom. The figures show that the transmit efficiency is higher for the directly fed dipole at the surface of the phantom, whereas the SAR efficiency is higher for the passively fed antenna for all depths. The reference point for the measurements was the feed point, ie the short feeding dipole was at the same distance from the phantom as the feed point of the directly fed dipole (note that in the case where the radiating conductors are instead used as reference points, the B 1 + efficiency increased by 2.7% at the surface and by less than 0.5% at 5 cm depth for the directly fed meander antenna).
In order to investigate the origin of the increased SAR efficiency,shows simulated surface current distribution on an antenna's conductor facing a phantom of (left) directly and (right) passively fed meander dipole.shows the simulated z-component of the E-field (E z ) on the surface of the dipoles. For both cases, the strongest E z values were at the feed point, around the meanders and at the end of the dipoles.shows the E z field from a side view of the simulated setup containing a directly/passively fed meander-dipole antenna placed on a 20 mm spacer above the phantom.shows the E z field at the surface of the phantom (top view). The E z field component was higher for the directly than for the passively fed dipole throughout the entire phantom.shows the measured S 11 of both dipole antennas as a function of the distance from the phantom, varying from 20 to 70 mm in steps of 10 mm. Both antennas were impedance-matched for a 20 mm distance from the phantom. As the distance increased, the S 11 minima of both antennas shifted towards higher frequencies by the same amount, indicating an equal amount of inductive coupling to the sample (note that the S 11 plot of the passively fed antenna has two peaks: one from the long passive element and the other from the short feeding dipole). shows the simulated B 1 + and B 1 − fields and SAR 10g, max values for the two different eight-channel arrays (both passively and directly fed antennas) on the body voxel model. The simulated B 1 + field was~12.5% lower in the prostate region for the passively fed meander-dipole arrays while the B 1 − field was~6% lower for the passively than the directly fed meander. The simulated SAR 10g, max produced by the passively fed dipole array is 44% lower than that from the directly fed dipole array, giving an increase in SAR efficiency of~15%.shows in vivo images from one of the four healthy volunteers. The experimentally (in vivo) measured noise correlation matrices of both arrays and for all four volunteers are shown in . The coupling coefficients between individual channels of both arrays were below −15 dB. The measured averaged B 1 + field in the prostate region was~15% lower, while the averaged SNR in a prostate was~5% lower for the array containing passively fed dipoles. shows measured noise correlation matrices of four volunteers with BMIs spanning from 20.8 to 28.7 kg/m 2 . In cases of both passively and directly fed arrays, the highest coupling coefficient was −15 dB. The highest inter-element coupling corresponded to the subject with the highest BMI value, which can be explained by the fact that tuning and matching of antennas were performed on a homogeneous phantom, where the antennas were placed at a 20 mm distance from the phantom, and for this particular volunteer the effective distance was much greater than this distance.
# | discussion
This study compares the transmit and SAR efficiencies of two identical meander-dipole configurations, the difference being that one is passively fed and the other is directly fed. The results show that the transmit efficiency of a passively fed dipole or dipole array is reduced by~15% at penetration depths of~10-15 cm, whereas the SAR 10g, max is reduced by more than 40% and the SAR efficiency is increased by~15%. Because max SAR 10g (normalized to 8 W of accepted power). Optimal phases of channels were determined for a maximum constructive B 1 + in prostate most high field body imaging applications are limited by the SAR rather than the absolute power available from the multiple RF amplifiers, these results suggest that changing from an actively fed to a passively fed architecture could have significant benefits for body imaging at high fields.
Why does a passive feed reduce the SAR and increase the SAR efficiency? In order to reduce SAR the total electric field penetrating the sample should be minimized, ideally without significantly affecting the magnetic B 1 + field component. The SAR depends on both the conservative and inductive (nonconservative) electric field components. The conservative electric field originates from a gradient in charge accumulation on the antenna's conductor. Fromand 4C, we conclude that the highest E z components (which can be attributed to the conservative electric field) occur at the location of the meanders, at the feed point, and at the end of the antenna's legs. The E z component is important because, based on the boundary conditions at the surface of the phantom, the tangential component of the electric field in the air and in the phantom must be the same. The main attribute of a passive-feed network is that it provides partial shielding of the E z component of an electric field in a driven device.shows that the strongest E z field of a directly fed antenna originates from the middle (feed) point of the dipole. In the case of the passively fed dipole, the middle point of the feed is "shielded" from the sample by the passive element and the induced E z field in the phantom is reduced. From the surface current distribution at the antenna's conductor facing the phantom, the maximum current distribution for the directly fed meander is in the central section of the antenna. The magnitude of the surface current distribution for the passively fed meander is lower in value and more evenly distributed along the conductor. Therefore, the inductive electric field, related to the surface current distribution on the antenna's radiating conductor, is lower for the passively fed meander. To confirm the hypothesis that the total electric field in a sample is a combination of both conservative and inductive components, we placed a small shielding strip below a conventional actively fed dipole.
We observed that the reductions in SAR were not as effective as those with the proposed passive-feed concept.
The inter-element coupling of the passively fed dipole (−15 to −18 dB) was found to be slightly less sensitive to in vivo loading conditions than that of the directly fed dipole (−20 to −28 dB), as reflected by the noise correlation coefficients shown in, although the absolute values were higher. Again, these results can be explained in terms of the partial shielding effect of the passive element, which reduces slightly the coupling to the sample.
The actual in vivo SAR value depends on the specific phases applied to the individual channels. In this study, we presented simulated SAR values of directly and passively fed dipole arrays where the phases are adjusted to produce the maximum constructive interference in a prostate region. One question to be addressed in future research is whether or not the directly fed dipole antennas would be able to reach a similar SAR efficiency with a different phase setting if some of the superior B 1 + efficiency were to be traded against avoiding local maxima in SAR.
# | conclusions
This study has shown that the SAR efficiency of a dipole element used for high field MRI can be increased by using a passive-rather than an active-feeding network. Although the results shown here are specifically for a meander dipole, the passive-feed network approach can also be applied to different dipole geometries. For example, we have performed simulations on the SAR efficiency of a passively fed straight dipole and found similar results to those presented here in terms of the improvement in SAR efficiency. We anticipate that similar behaviour would be found for other types of dipole such as folded,snake,or the recently introduced dual mode.We also anticipate that a passive-feeding network would increase the SAR efficiency of electrically short dipoles, which are known to have high transmit efficiency but relatively low SAR efficiency, and are suitable, for example, for cardiac or pancreatic imaging. |
Metallothioneins (MT) are low molecular weight, cysteine-rich proteins maintaining metal ions homeostasis. They play a role in carcinogenesis and may also cause chemoresistance. The aim of the study was to explore the importance of MT serum levels in children suffering from malignant tumours. This prospective study involves examination of 865 samples from 172 patients with malignant tumours treated from 2008 to 2011 at University Hospital Motol. MT serum levels were determined using differential pulse voltammetry-Brdicka reaction. Mean MT level was 2.7 ± 0.5 μM. There was no statistically significant difference between MT levels in different tumours. We also did not find any correlation between MT levels and response to therapy or clinical stages. However, we found a positive correlation between MT levels and age (p = 0.009) and a negative correlation with absolute lymphocyte number (p = 0.001). The fact that patients who had early disease recurrence had lower MT levels during the treatment (complete OPEN ACCESS Int. J. Mol. Sci. 2013, 14 12171 remission 2.67 vs. recurring 2.34, p = 0.001) seems to be important for clinical practice. Accordingly we believe that there is benefit in further studies of serum MT levels in tumours.
enhancers in the MT-2A promoter region [bib_ref] Molecular analyses of metallothionein gene regulation, Samson [/bib_ref]. However, the information on serum levels of individual isoforms is still lacking.
The aim of this study is to explore relationships between MT serum levels and type of tumour, stages and response to therapy in children suffering from solid tumours. We also were looking for correlation between MT levels and other laboratory findings.
# Results
## Mt in serum
The mean MT level in children with malignant tumours was 2.67 ± 0.5 μM. There were no statistically significant differences between different tumours in our group of patients [fig_ref] Figure 1: MT level/mean ± s [/fig_ref]. The highest mean MT levels were found in germ cell tumours (MTs 2.94 ± 0.9 μM) and the second highest levels were determined in brain tumours (2.82 ± 0.5 μM) but the differences are not statistically significant and are probably caused by age distribution. We previously published average MT levels in serum of healthy adult volunteers detected also using the Brdicka method. In the sample (mean age 27 years), MT level was 0.52 ± 0.2 μM [bib_ref] Femtogram electroanalytical detection of prostatic specific antigen by brdicka reaction, Hynek [/bib_ref]. Milnerowicz and Bizoń found similar results in healthy adult volunteers using enzyme-linked immunosorbent assay [bib_ref] Determination of metallothionein in biological fluids using enzyme-linked immunoassay with commercial antibody, Milnerowicz [/bib_ref] and Singh and Hanson found lower MT levels in 50 normal children (aged 2-11 years) by immunoassay (approx. 2 mg/L) [bib_ref] Assessment of metallothionein and antibodies to metallothionein in normal and autistic children..., Singh [/bib_ref]. Somewhat higher were MT levels detected by differential pulse voltammetry-Brdicka reaction found in a group of children (age 3-17 years) with various renal diseases with normal creatinine levels (MTs 1.42 ± 0.09 μM). Those levels in healthy adult controls are approximately five times lower than the average (2.67 ± 0.5 μM) MT levels found in our group of patients. The MT levels in the group of children with renal diseases were roughly half that of values in children suffering from malignant tumours. Both those differences were statistically significant (T-test, p < 0.01). The MT levels found in a group of children with nonmalignant kidney diseases were approximately half of levels in our group with malignant tumours. We found highly significant positive correlation between MT levels and age that was independent of tumour type. [fig_ref] Figure 1: MT level/mean ± s [/fig_ref] depicts a general linear model with dependent variable MT level, between subjects factor type of tumour and covariate age, displaying significance of covariate effect age p = 0.009 (F = 7.05, df = 1), parameter estimate of age B = 0.023 (positive).
## Mt relationship to course of the disease
We did not find any differences in MT levels among patients with tumours before starting therapy, with tumour and/or metastases during therapy or with progressive disease (group of "patients with active disease") and patients in complete remission (group without "active disease"). There were no statistically significant differences in MT levels in patients with metastatic diseases in comparison with levels in patients with localized cancer. The majority patients displaying lower MTs levels had early recurrence of disease. Average MT levels during all follow up period before detection of recurrence was 2.34 μM, versus complete remission 2.67 μM (p = 0.001). These results are shown in [fig_ref] Figure 2: Relation of MT level/mean ± s [/fig_ref]. We, however, found significant decrease of MT levels in children during follow up ≥2 months after the end of the treatment as compared with levels at the end of therapy (p = 0.001, [fig_ref] Figure 3: MT levels/mean ± s [/fig_ref].
## Mt levels and chemotherapy
MT levels in patients with active disease were higher in those treated with chemotherapy than without (p = 0.02). There was no significant difference in MT levels in patients treated by protocols containing platinum cytostatics and treated without those cytostatics [fig_ref] Figure 4: MT levels in relation to chemotherapy [/fig_ref]. : not significant. p = 0.02: statistically significant on the level of 2%. The only significant difference was found between levels in patients with active tumour (before therapy, partial remission, progressive disease) with and without chemotherapy. There was no relationship to chemotherapy with platinum cytostatics.
## Correlation of mt levels with other biochemical parameters
There was a negative correlation between MT levels and absolute lymphocyte count (p = 0.001) and positive correlation between MT and creatinine levels (p = 0.003). Correlation between haemoglobin and MT levels was below the level of statistical significance (p = 0.085). Patients with CRP above normal levels exhibited no difference in MT level compared to those with normal CRP (p = 0.912). Summary of the results obtained is shown in [fig_ref] Table 1: Correlations between laboratory parameters [/fig_ref]. Correlations of MT levels with other laboratory findings mentioned in Materials and methods were not significant.
# Discussion
Numerous studies have been carried out to understand the relationship between MT in cells and cancer. Most of these have focused on adults [bib_ref] Mammalian metallothionein in toxicology, cancer, and cancer chemotherapy, Namdarghanbari [/bib_ref] [bib_ref] Molecular functions of metallothionein and its role in hematological malignancies, Takahashi [/bib_ref] [bib_ref] Interaction of metallothionein with tumor suppressor p53 protein, Ostrakhovitch [/bib_ref] [bib_ref] Indicators of zinc status at the population level: A review of the..., Gibson [/bib_ref] [bib_ref] The possible biological role of metallothionein in apoptosis, Dutsch-Wicherek [/bib_ref] [bib_ref] Identification of bladder cancer patients at risk for recurrence or progression: An..., Hinkel [/bib_ref] [bib_ref] Expression of metallothionein and p53 antigens are correlated in oral squamous cell..., Cardoso [/bib_ref] and far fewer on children [bib_ref] bcl-2. P21(waf1/cip1) and metallothionein as markers of differentiation, response to treatment and..., Lara-Bohorquez [/bib_ref] [bib_ref] Portrait of ependymoma recurrence in children: Biomarkers of tumor progression identified by..., Peyre [/bib_ref] [bib_ref] Osteosarcoma is characterised by reduced expression of markers of osteoclastogenesis and antigen..., Endo-Munoz [/bib_ref] [bib_ref] Expression of metallothionein in initial and relapsed childhood acute lymphoblastic-leukemia, Sauerbrey [/bib_ref]. Studies in adults have shown that cellular MT could serve as prognostic markers in some tumours [bib_ref] Mammalian metallothioneins: Properties and functions, Babula [/bib_ref] [bib_ref] Metallothionein-Overexpression as a highly significant prognostic factor in melanoma: A prospective study..., Weinlich [/bib_ref] [bib_ref] Metallothionein-overexpression as a prognostic marker in melanoma, Weinlich [/bib_ref] [bib_ref] Combined analysis of hypoxia-inducible factor 1 alpha and metallothionein indicates an aggressive..., Schmitz [/bib_ref] [bib_ref] Metallothionein and p-akt proteins in oral dysplasia and in oral squamous cell..., Pontes [/bib_ref] [bib_ref] Over-expression of metallothionein predicts chemoresistance in breast cancer, Yap [/bib_ref] [bib_ref] Overexpression of the drug resistance-associated protein metallothionein does not correlate with response..., Grabellus [/bib_ref]. Relationships have been described between MT and p53 status [bib_ref] Metallothioneins and cancer, Eckschlager [/bib_ref] , increased tumour grade [bib_ref] Zinc, antioxidant systems and metallothionein in metal mediated-apoptosis: Biochemical and cytochemical aspects, Formigare [/bib_ref] [bib_ref] Immunohistochemical analysis of metallothionein in astrocytic tumors in relation to tumor grade,..., Hiura [/bib_ref] and metastases development [bib_ref] Metallothioneins and cancer, Eckschlager [/bib_ref] [bib_ref] Correlation of metallothionein expression with clinical progression of cancer in the oral..., Szelachowska [/bib_ref]. However what has been proved in adults is not so clear in childhood cancers. There is one study in children with osteosarcoma, where the authors found differential expression of MT in biopsies [bib_ref] Osteosarcoma is characterised by reduced expression of markers of osteoclastogenesis and antigen..., Endo-Munoz [/bib_ref] and they suggested that MT might play an important role in development of disease. Another study on children suffering from leukaemia showed that patients with acute lymphoblastic leukaemia whose blasts express MT tended to have shorter disease-free survival compared with the MT negative ones [bib_ref] Expression of metallothionein in initial and relapsed childhood acute lymphoblastic-leukemia, Sauerbrey [/bib_ref]. A further report on patients with neuroblastoma demonstrated only limited value of MT for prediction of therapeutic response [bib_ref] bcl-2. P21(waf1/cip1) and metallothionein as markers of differentiation, response to treatment and..., Lara-Bohorquez [/bib_ref].
MT levels in serum have seldom been researched in cancer patients. They have been used for monitoring intoxication by metals (lead, cadmium) [bib_ref] Enzyme-immunoassay for the determination of metallothionein in human urine: Application to environmental..., Swierzcek [/bib_ref]. Previously, our group has published three publications on MT in serum in adult cancer. Increased MT serum levels in prostate cancer patients were found to be negatively correlated with their Gleason score [bib_ref] Evaluation of alpha-methylacyl-CoA racemase, metallothionein and prostate specific antigen as prostate cancer..., Gumulec [/bib_ref]. MT levels in patients suffering from head and neck spinocellular cancer demonstrated correlation with tumour grade and clinical stage [bib_ref] Metallothionein electrochemically determined using brdicka reaction as a promising blood marker of..., Krejcova [/bib_ref] [bib_ref] Study of metallothionein role in spinocellular carcinoma tissues of head and neck..., Sochor [/bib_ref]. In another study hepatocellular carcinoma serum MT levels were decreased compared to the control group [bib_ref] A new diagnostic method for chronic hepatitis, liver cirrhosis, and hepatocellular carcinoma..., Nakayama [/bib_ref]. To our knowledge no work has yet been published on MT serum levels in child cancer patients.
Our study covered a wide range of child cancers. MTs were found not to act as a prognostic marker of treatment response in any of these diseases. Patients with metastases or progressive disease did not exhibit different MT levels either. Differences between different types of tumours were not significant. The highest mean MT levels we found were in germ cell tumours (MT 2.94 ± 0.9 μmol·dm −3 ) most likely because these included a higher proportion of older patients. The second highest were in brain tumours (2.82 ± 0.5 μmol·dm −3 )-probably because the brain, in general, contains higher amounts of MT [bib_ref] Metallothionein reduces central nervous system inflammation, neuro degeneration, and cell death following..., Penkowa [/bib_ref] [bib_ref] Prognosis-related histomorphological and immunohistochemical markers in central nervous system tumors of childhood..., Rickert [/bib_ref] [bib_ref] Metallothionein i and ii mitigate age-dependent secondary brain injury, Natale [/bib_ref] , but age distribution may also play a role. During anticancer therapy there are many different pathological reactions (tissue damage induced by cytostatics and/or radiotherapy, supportive therapy including antibiotics and antimycotics, erythropoetin and/or filgrastim, infections, tumour behaviour, etc.). Each of these contributes to various and hence continuously varying levels of oxidative stress. MTs are known to function as one of the main antioxidant defence systems [bib_ref] Metallothioneins and cancer, Eckschlager [/bib_ref] [bib_ref] The metallothionein/thionein system: An oxidoreductive metabolic zinc link, Bell [/bib_ref] [bib_ref] Biomarkers for oxidative stress: Clinical application in pediatric medicine, Tsukahara [/bib_ref]. Changes of MT levels caused by therapy and/or its complications mentioned above are probably the main reason why correlation between tumour activity and MT levels is not significant. We have not found significant correlation between tumour activity and MT levels.
The finding that low MT levels during treatment were closely correlated with early relapses may be important for clinical practice. Peyere et al. found that up to 80% of children with recurrence of ependymoma have down-regulated MT-3 genes, not caused by MT gene deletion or promoter methylation. MT-3 expression was restored by histondeacetylase inhibitor or zinc treatment [bib_ref] Portrait of ependymoma recurrence in children: Biomarkers of tumor progression identified by..., Peyre [/bib_ref]. One may suggest that down-regulation could be a key cause of low MT levels in patients suffering from recurrence, although from twelve recurrent tumours, only three were brain tumours and it is not known if other MT isoforms expression is also down regulated in recurrent tumours of other organs.
The explanation of slow decline (2 months) of MT levels after completing therapy is long recovery reactions in the organisms after the end of treatment. We choose a two month limit because it is an interval in which haematopoiesis and immunity is usually reconstituted after conventional doses of chemotherapy, and one may speculate that other systems are also normalized during this interval [bib_ref] Chemotherapy-induced immunosupression, Arvin [/bib_ref]. This suggests that chemotherapy influences MT levels by multiple mechanisms e.g. the damage of different organs and systems, the increasing risk of infections, and/or by the production of reactive oxygen species. It will be necessary to observe these patients over longer periods to determine when MT levels return to normal. We suggest that changes of MT levels could be considered as a promising marker for late relapses. In our study we did not have patients who suffer late recurrence more than 2 years after the end of treatment. The fact that MT levels declined during the two months following the end of treatment supports our view that the anticancer and supportive therapy significantly affects the serum levels of MT.
MT plays an important role in chemoresistance and not only to metal containing drugs. Their expression may be increased not only by metals but by other stimuli e.g. glucocorticoids, catecholamines, free radicals, tumour necrosis factor α, interleukins-1, -2 and -6 [bib_ref] Metallothioneins and cancer, Eckschlager [/bib_ref] [bib_ref] Mammalian metallothioneins: Properties and functions, Babula [/bib_ref]. Production of some those MT-inducing biologically active molecules is stimulated by chemotherapy. This may explain increased MT levels in patients with "active tumours" treated with chemotherapy.
The findings of increased MT levels in patients with active disease (with tumours before starting therapy, with tumour and/or metastases during therapy or with progressive disease) but not in remission treated by cytostatics may be explained by the production of MT by cancer cells stimulated by chemotherapy. MTs are able to bind platinum-based cytostatics and thereby reduce their cytotoxic effect. Chemoresistance to platinum anti-tumour compounds is mediated through several mechanisms. One of them is the transfer of platinum from cisplatin and carboplatin to MT that results in inactivation of those drugs. Cultivation of neuroblastoma cells resistant to cisplatin in medium with cisplatin or carboplatin, has been shown to significantly increase intracellular MT levels [bib_ref] Metallothioneins and cancer, Eckschlager [/bib_ref]. However in a sensitive cell line only insignificant increases in MT were detected after cultivation with the same concentrations of cisplatin or carboplatin [bib_ref] Metallothioneins and cancer, Eckschlager [/bib_ref]. Another study showed that a cisplatin-resistant ovarian cancer cell line exposed to cisplatin manifested a nuclear MT expression [bib_ref] Nuclear metallothionein expression correlates with cisplatin resistance of ovarian cancer cells and..., Surowiak [/bib_ref]. In hepatoblastoma patients treated with carboplatin, it was verified that non-responders had a higher percentage of MT-positive tumour cells [bib_ref] Immunohistochemical metallothionein expression in hepatocellular carcinoma: Relation to tumor progression and chemoresistance..., Endo [/bib_ref]. The significance of MT expression to resistance of gastric cancer to cisplatin was verified by Suganuma et al. [bib_ref] Possible chemoresistance-related genes for gastric cancer detected by cdna microarray, Suganuma [/bib_ref]. MT upregulation was detected in medulloblastoma and rhabdomyosarcoma cells with induced resistance to the alkylating drug BCNU [bib_ref] Mechanisms of resistance to 1,3-bis(2-chloroethyl)-1-nitrosourea in human medulloblastoma and rhabdomyosarcoma, Bacolod [/bib_ref]. Esophageal carcinomas which do not express MT, respond well to chemoradiotherapy (5-fluorouracil and cisplatin) while cancers with high MT expression are resistant [bib_ref] P53 negativity, cdc25b positivity, and metallothionein negativity are predictors of a response..., Sunada [/bib_ref]. Women with breast carcinoma treated with chemotherapy (cyclophosphamide, methotrexate, 5 fluorouracil or doxorubicin) had significantly longer survival if their tumours had lower MT expression [bib_ref] Over-expression of metallothionein predicts chemoresistance in breast cancer, Yap [/bib_ref]. Hishikawa and co-workers found that MT negative patients with oesophageal cancer treated with cisplatin experienced increased survival compared to those with MT positive tumours [bib_ref] Overexpression of metallothionein correlates with chemoresistance to cisplatin and prognosis in esophageal..., Hishikawa [/bib_ref]. It has been found that MT expression initially increases with age, but then decreases in people over seventy [bib_ref] Metallothionein downregulation in very old age: A phenomenon associated with cellular senescence?, Malavolta [/bib_ref] and the inducibility of MT increase with age in infant rat model was described by Bauerly [bib_ref] Effects of copper supplementation on copper absorption, tissue distribution, and copper transporter..., Bauerly [/bib_ref]. Correlation of serum MT levels found in our study may also indirectly confirm this phenomenon. Natale et al. showed in animal experiments that low MT explains enhanced susceptibility to neuronal loss after injury in immature brains [bib_ref] Metallothionein i and ii mitigate age-dependent secondary brain injury, Natale [/bib_ref]. It is common knowledge that chemotherapy is more toxic in very young children compared to older children. Our findings of low MT in infant patients could be one of the possible reasons for more serious side effects of chemotherapy and radiotherapy in infants compared to older children.
The possible explanation of decreased MT in patients with higher creatinine levels may be kidney damage, as they are one of the major producers of MTs. MT-1 and MT-2 are produced particularly in kidney, liver, pancreas and intestine [bib_ref] Metallothioneins and cancer, Eckschlager [/bib_ref]. The positive correlation of MT levels and lymphocyte count in cancer patients do not have a clear explanation. Due to significant correlations between serum creatinine and lymphocyte counts and haemoglobin levels, it is difficult to assess whether the relationship between MT and creatinine is primary, or whether there are primary relationships for haematological parameters. It is possible that the induction of MT by cytokines such as interleukin-1, -2 and -6 or tumour necrosis factor, which are produced by lymphocytes, may play a possible role.
It has been found that zinc supplementation influences lymphocyte production [bib_ref] Metallothionein downregulation in very old age: A phenomenon associated with cellular senescence?, Malavolta [/bib_ref] [bib_ref] Genome-level expression profiles in pediatric septic shock indicate a role for altered..., Wong [/bib_ref]. The authors looked at changes in large numbers of genes that were involved in zinc homeostasis in peripheral blood leukocytes of children with septic shock. MT expression in children who died was also increased. It was suggested that decline in zinc concentrations among critically ill children was related to shifts in MT expression and low plasma zinc levels were associated with the degree of organ failure [bib_ref] Effects of copper supplementation on copper absorption, tissue distribution, and copper transporter..., Bauerly [/bib_ref] [bib_ref] Genome-level expression profiles in pediatric septic shock indicate a role for altered..., Wong [/bib_ref] [bib_ref] Zinc homeostasis in pediatric critical illness, Cvijanovich [/bib_ref]. Child patients who have low lymphocyte count also have increased risk of sepsis. In our study we noticed that patients with low lymphocyte count had increased MT levels. Three of our patients had life threatening sepsis during leucopoenia and high serum MT levels-2.1 times higher than mean (data not shown), which is in agreement with the studies mentioned above. The fact that CRP did not correlate with MT levels suggests that MTs are not acute-phase proteins, and may be a laboratory marker that is independent of acute-phase.
## Experimental section
## Patients
This prospective study involves examination of 865 samples from 172 patients with malignant tumours treated at the Department for Paediatric Haematology and Oncology, University Hospital Motol from 2008 to 2011. Samples were collected before starting chemotherapy, during chemotherapy and after. Diagnoses and age distribution see [fig_ref] Table 2: Clinical characteristics of patients group [/fig_ref]. The group included 71 girls (41%) and 101 boys (59%). Other clinical parameters were: metastatic disease 93 (54.1%), death during follow up 32 (18.6%), recurrence during follow up 12 (7%). The minimal follow up was 18 months and median follow up was 39 m [fig_ref] Table 2: Clinical characteristics of patients group [/fig_ref]. In our study we investigated the correlation between MT levels and clinical parameters (age, diagnosis, clinical stage, recurrence and response to anticancer therapy). We also compared MT levels examined before, during and after therapy with other laboratory findings: total blood count and urea, creatinine, uric acid, lactate dehydrogenase, transaminases, bilirubin, ferritin, total protein, and C-reactive protein (CRP) determination. Total blood count was examined at the Department of Clinical Haematology, and other biochemical tests at Department of Clinical Biochemistry and Pathobiochemistry, University Hospital Motol, according to standard protocols.
## Determination of metallothioneins
Samples were prepared by heat treatment using an automated pipetting system epMotion 5075 (Eppendorf, Hamburg, Germany) and kept at 4 °C then transferred to the 96 well plates (Eppendorf, Hamburg, Germany) together with 0.2 M phosphate buffer pH 7. This mixture was kept at 99 °C for 15 min. The last step was cooling down of samples to 4 °C. Heating denatures and removes the high molecular weight proteins from samples [bib_ref] Evaluation of different purification procedures for the electrochemical quantification of mussel metallothioneins, Erk [/bib_ref]. MT quantification was determined by electrochemical detection. Differential pulse voltammetric Brdicka reaction measurements were performed using a 747 VA Stand instrument connected to a 693 VA Processor and 695 Autosampler (Metrohm, Herisau, Switzerland), using a standard cell with three electrodes, a cooled sample holder and measurement cell. Measurements were taken on samples cooled to 4 °C (Julabo F25, Julabo, Seelbach, Germany). A hanging mercury drop electrode (HMDE) with a drop area of 0.4 mm 2 was the working electrode. An Ag/AgCl/3M KCl electrode was the reference and platinum electrode was auxiliary. For data processing VA Database 2.2 by Metrohm was employed. The analysed samples were deoxygenated prior to measurements by purging with argon (99.999%) saturated with water for 120 s. A Brdicka supporting electrolyte containing 1 mM Co(NH 3 ) 6 Cl 3 and 1M ammonia buffer pH = 9.6 was used. The supporting electrolyte was exchanged after each measurement. The parameters of the measurement were as follows: initial potential of −0.7 V, end potential of −1.75 V, modulation time 0.057 s, time interval 0.2 s, step potential 2 mV, modulation amplitude −250 mV, E ads = 0 V, volume of injected sample: 10 µL, volume of measurement cell 2 mL (5 μL of sample + 1995 µL Brdicka solution) [bib_ref] Evaluation of alpha-methylacyl-CoA racemase, metallothionein and prostate specific antigen as prostate cancer..., Gumulec [/bib_ref]. Electrochemical determination of metallothionein using Brdicka reaction is described in detail in our previously published papers [bib_ref] Femtogram electroanalytical detection of prostatic specific antigen by brdicka reaction, Hynek [/bib_ref] [bib_ref] Study of metallothionein role in spinocellular carcinoma tissues of head and neck..., Sochor [/bib_ref] [bib_ref] Analytical methods for metallothionein detection, Ryvolova [/bib_ref] [bib_ref] Analysis of metallothionein by capillary electrophoresis, Ryvolova [/bib_ref] [bib_ref] Employment of electrochemical techniques for metallothionein determination in tumor cell lines and..., Fabrik [/bib_ref] [bib_ref] An electrochemical detection of metallothioneins at the zeptomole level in nanolitre volumes, Adam [/bib_ref]. In a Brdicka solution, MT gives four very well separated signals [bib_ref] Elucidation of the mechanism of the brdicka reaction, Raspor [/bib_ref] [bib_ref] Zeptomole electrochemical detection of metallothioneins, Adam [/bib_ref] : 1/Cat1 (potential app. −1.25 V); 2/Cat2 (potential app. −1.45 V) signals correspond to the reduction of hydrogen at the mercury electrode (hydrogen is generated from the supporting electrolyte; 3/RS 2 Co (potential app. −1.1 V) signal represents response of MT complex to components of Brdicka's supporting electrolyte; 4/ the signal called Co1, at the potential app. −1.0 V, probably relates to the reduction of the RS 2 Co complex. The first three MT signals of Brdicka reaction (RS 2 Co, Cat1 and Cat2) displayed well developed and separated peaks with decreasing MT concentration. Signal Co1 decreased and shifted to more negative potential with decreasing MT [bib_ref] Zeptomole electrochemical detection of metallothioneins, Adam [/bib_ref].
# Statistical analysis
Data were evaluated as parametrical. Kolmogorov-Smirnov test was used for testing of normal distribution of MT levels and it was not rejected (K-S test, p = 0.17). Data were expressed as mean ± standard deviation. Differences p < 0.05 were considered as significant. Used tests: for comparison of MT levels between two groups independent t-test, in paired data paired samples t-test, general linear model with 1 between subjects factor and 1 covariate for evaluation of influence of factor with account of age, for correlation of MT levels with other parameters Pearson correlation coefficient or partial correlation coefficient with controlling variable age (correlation of MT with creatinine, because both are growing with age) and for comparison among groups One way ANOVA with Post hoc tests Scheffe was used. IBM SPSS Statistics (Release 20.0, IBM Corporation, Armonk, NY, USA) was used.
# Conclusions
We found that differential pulse voltammetry with automatic sample preparation is suitable for quantification of MT in serum. This method is fast, relatively inexpensive and enables efficient analysis of large numbers of samples, but it is so far unable distinguish different MT isoforms. In spite of the fact that the Brdicka reaction is not sensitive to the single isoforms, specificity of the sample preparation and isolation is higher than 90% [bib_ref] Zeptomole electrochemical detection of metallothioneins, Adam [/bib_ref] [bib_ref] Using of chicken antibodies for metallothionein detection in human blood serum and..., Krizkova [/bib_ref]. There are increased serum MT levels in children suffering from cancer compared to healthy volunteers described in our previous study and a group of children with kidney diseases and also from groups of healthy persons described in literature and their serum MT levels decline after finishing chemotherapy. We found a correlation between MT levels and age and absolute lymphocyte number in our group of patients. There was no correlation with tumour responses and stages. All patients with early relapses had low MT levels. Hence, one may speculate that lower MT levels could be a promising marker for early relapses. Accordingly we believe that there is benefit in further larger studies of serum MT levels in child cancer patients. Further studies should be extended to the examination of individual MT isoforms since they play different functional roles.
[fig] Figure 1: MT level/mean ± s.d./in different childhood malignant tumours. Hatched column: all tumours; doted columns: healthy adult volunteers. There were no statistically significant differences between different tumours, difference between MT levels in all malignant tumours and in healthy volunteers was significant. MMT: Soft tissue sarcoma (Malignant mesenchymal tumours). [/fig]
[fig] Figure 2: Relation of MT level/mean ± s.d./to clinical parameters. N.S.: not significant. p = 0.001: statistically significant on the level of 0.1%. There were no statistically significant differences between MT levels in patients with active disease and in remission or in patients with generalized disease and with cancers without metastasis. [/fig]
[fig] Figure 3: MT levels/mean ± s.d./in group of patients <2 months after finishing chemotherapy compared to patients ≥2 months after finishing chemotherapy [paired T-test (t = 3.53, df = 75) p = 0.001]. [/fig]
[fig] Figure 4: MT levels in relation to chemotherapy. N.S. [/fig]
[table] Table 1: Correlations between laboratory parameters.Table 1. Cont. [/table]
[table] Table 2: Clinical characteristics of patients group. * spinal metastases, ** no of treated by any cytostatic/no of treated by Pt containing cytostatic (cisplatin and/or carboplatin). [/table]
|
Prognostic nomograms for patients with resectable hepatocelluar carcinoma incorporating systemic inflammation and tumor characteristics
Background: The model to predict the prognosis of resectable hepatocelluar carcinoma (HCC) has not been determined. Methods: Predictors were selected using Cox model. Nomograms were generated in the training set and validated in the validation set. The predictive ability of the nomogram was determined by concordance index and calibration curve. Results: Independent factors for overall survival including alpha-fetoprotein level (hazard ratio (HR):1.292), tumor size (HR:1.092), tumor number (HR:1.472), microvascular invasion (HR:1.660), neutrophil to lymphocyte count ratio (NLR) (HR:1.428), major vascular invasion (HR:2.485) and satellite lesions(HR:1.392) were selected into the nomogram for survival. The c-index in the training set and validation set were 0.767 and 0.719, respectively, which were statistically higher than those of the four conventional staging systems.(Barcelona Clinic Liver Cancer: 0.644 and 0.609; the seventh American Joint Committee on Cancer: 0.678 and 0.674; Cancer of the Liver Italian Program: 0.692 and 0.648; Hong Kong Liver Cancer: 0.689 and 0.639, p < 0.001 for all). A nomogram for predicting 3-and 5-year recurrence free survival was generated with the c-index of 0.746 for the training set and 0.718 for the validation set, respectively. Conclusions: We have generated nomograms predicting prognosis for HCC treated by hepatectomy with a higher predictive power.
# Introduction
Hepatocelluar carcinoma (HCC) in china alone accounts for appropriate 50% of the deaths from liver cancer worldwide due to high presence of Hepatitis B virus (HBV) infection. [bib_ref] The global burden of liver disease: the major impact of China, Wang [/bib_ref] Surgical resection is widely accepted for curative therapy. Small HCC with well preserved liver function is suitable for hepatectomy with 5-year survival rate of 70-81%. [bib_ref] Surgical Resection Versus Radiofrequency Ablation for Single Hepatocellular Carcinoma </ = 2..., Liu [/bib_ref] [bib_ref] Postoperative prognostic nutritional index change is an independent predictor of survival in..., Peng [/bib_ref] For patients with intermediate-advanced stage HCC, hepatectomy is not recommended based on current stage systems due to high recurrence rate, recent studies from western and eastern shown that surgical resection for some selected patients with intermediated-advanced stage HCCs could bring more survival benefits than transarterial chemoembolization (TACE). [bib_ref] And Italian Liver Cancer g. Survival benefit of liver resection for patients..., Vitale [/bib_ref] [bib_ref] Hepatic resection associated with good survival for selected patients with intermediate and..., Zhong [/bib_ref] [bib_ref] Surgical treatment of hepatocellular carcinoma associated with hepatic vein tumor thrombosis, Kokudo [/bib_ref] With more patients suitable for hepatectomy, the model to predict all resectable HCCs but not only early stage HCC or specific stage HCCs was necessary. [bib_ref] Nomograms for Pre-and Postoperative Prediction of Longterm Survival for Patients Who Underwent..., Yang [/bib_ref] [bib_ref] Prognostic nomograms for prediction of recurrence and survival after curative liver resection..., Shim [/bib_ref] Liu et al. had proposed one model which had excellent discriminating power about prognosis of patients from early to advanced stages of HCC. [bib_ref] Proposal and validation of a new model to estimate survival for hepatocellular..., Liu [/bib_ref] We believed that, for resectable, HCCs, the prediction power of nomograms with full consideration of pathological findings or/and some promising inflammation index might increase.
The conventional staging systems mainly focused on tumor status, such as tumor size, tumor number and vascular invasion.They had some limitation in predicting the prognosis of HCC after surgery because the current staging systems put insufficient attention on pathological findings, such as microvascular invasion (MVI), satellite lesions, and acceptable risk factors like
## Research paper
Oncotarget 80784 www.impactjournals.com/oncotarget platelet-to-lymphocyte ratio (PLR) or/and neutrophil-tolymphocyte ratio (NLR). For example, MVI representing tumor behavior had been demonstrated as an extremely important factor associated with the prognosis of HCCs. [bib_ref] Microvascular invasion is a better predictor of tumor recurrence and overall survival..., Lim [/bib_ref] [bib_ref] Milan criteria, multinodularity, and microvascular invasion predict the recurrence patterns of hepatocellular..., Hung [/bib_ref] Only the seventh American Joint Committee on Cancer (AJCC) system had included this factor to distinguish the prognosis. Similarly, satellite lesion negatively impacting the prognosis was neglected in the conventional staging systems. [bib_ref] Treatment of early hepatocellular carcinoma: How to predict and prevent recurrence, Andreana [/bib_ref] [bib_ref] Long-term outcomes and prognostic factors of elderly patients with hepatocellular carcinoma undergoing..., Huang [/bib_ref]. Besides, in most investigations, tumor size itself as continuous variable was included as a categorical variable and tumor number were simply classified as single and multiple HCCs. [bib_ref] Tumor size of hepatocellular carcinoma in noncirrhotic liver: a controversial predictive factor..., Truant [/bib_ref] [bib_ref] The Impact of Tumor Size on Long-Term Survival Outcomes After Resection of..., Hwang [/bib_ref] [bib_ref] Effect of Tumor Size on Cancer-Specific Survival in Small Hepatocellular Carcinoma, Zhang [/bib_ref] [bib_ref] A nomogram predicting pulmonary metastasis of hepatocellular carcinoma following partial hepatectomy, Li [/bib_ref] Previous studies showed patients with tumor number more than 3 might present poor prognosis compared to patients with 1-3 tumors after surgery. [bib_ref] Barcelona clinic liver cancer stage B hepatocellular carcinoma: transarterial chemoembolization or hepatic..., Jianyong [/bib_ref] We hypothesized model incorporating tumor size as continuous variable might accurately predict the prognosis since tumor size was closely related with risk factors, such as MVI and satellite lesion. [bib_ref] Tumor size predicts vascular invasion and histologic grade: Implications for selection of..., Pawlik [/bib_ref].
Recently, systematic inflammation index, such as PLR and NLR, has been focused by many investigators and demonstrated to be an independent risk factor associated with cancerous prognosis. [bib_ref] Neutrophil to lymphocyte ratio as an indicator of the malignant behaviour of..., Okamura [/bib_ref] [bib_ref] Prognostic role of platelet to lymphocyte ratio in non-small cell lung cancers:..., Zhao [/bib_ref] [bib_ref] Platelet-to-lymphocyte ratio acts as a prognostic factor for patients with advanced hepatocellular..., Li [/bib_ref] Inflammation closely correlated with cancer progress [bib_ref] Contribution of platelets to tumour metastasis, Gay [/bib_ref] [bib_ref] Cancer: Inflammation lights the way to metastasis, Coffelt [/bib_ref] In clinical, neutrophil counts, platelet counts and lymphocyte counts might reflect the host immune and inflammation status and thus indicate tumor progress through different combinations. [bib_ref] Neutrophil to lymphocyte ratio as an indicator of the malignant behaviour of..., Okamura [/bib_ref] Unfortunately, they are few predictive models including these simple and effective factors.
Nomogram is a method that could assign relative risk scores to each predictor according to its contribution to the prognosis. Based on total points of individual, it could predict the probability of overall survival (OS) and recurrence free survival (RFS) after hepatectomy. Recently, many investigators successfully evaluate risk scores to more accurately predict the outcomes in various cancers including HCC. [bib_ref] Conditional probability of survival nomogram for 1-, 2-, and 3-year survivors after..., Dikken [/bib_ref] [bib_ref] A Nomogram to Predict Recurrence and Survival of High-Risk Patients Undergoing Sublobar..., Kent [/bib_ref] [bib_ref] An accurate prognostic staging system for hepatocellular carcinoma patients after curative hepatectomy, Tokumitsu [/bib_ref]. For resectable HCCs after hepatectomy, model incorporating all the available risk factors might increase power in predicting the prognosis of HCCs after hepatectomy.
The purpose of the current study is (1) to construct clinically useful nomograms with full consideration of tumor status and systematic inflammation to predict OS and RFS for resectable HCC after hepatectomy; (2) to validate the nomograms in the validation set; (3) to compare its predictive power with current popular staging systems.
# Results
## Baseline characteristics in training set and validation set
A total of 777 subjects were included in the current study with 618 in the training set and 159 subjects in the validation set. The baseline characteristics of the two cohorts in all study subjects are shown in
## Follow-up treatments and predictors of os
## Construction and validation of the nomogram
A nomogram to predict OS of patients with HCC following surgical resection is shown in [fig_ref] Figure 1: Nomograms to predict probability of 3-and 5-year over survival [/fig_ref]. Spearman correlation indicated that surgery type versus tumor size correlated significantly (r = 0.402, P < 0.001). Except for surgery type, the nomogram included other seven variables associated with OS in multivariable analysis. Each factor in the nomogram was assigned a weighted number of points, and the sum of points for each patient was in accordance with a specific predicted 3-and 5-year OS. A higher score predicted worse prognosis [fig_ref] Figure 1: Nomograms to predict probability of 3-and 5-year over survival [/fig_ref].The model demonstrated good accuracy for predicting overall survival rate of HCC after hepatectomy, with a c-index of 0.767 (95% CI 0.742-0.792). The bootstrapped calibration plot for the prediction of 3-year and 5-year OS is shown in [fig_ref] Figure 2: Good calibration of the nomogram in the training and validation set [/fig_ref] and 2B. The calibration plots revealed good prediction of 3-and 5-year OS.
In the validation set, the C-index of the nomogram for predicting OS was 0.719 (95% CI, 0.671 to 0.767), and a calibration curve showed good agreement between Similarly, based on the multivariate analysis, the seven risk factors included in the OS nomogram were indentified predictors of recurrence free survival [fig_ref] Table 3: Variables associated with RFS according to the Cox proportional hazards model [/fig_ref]. A nomogram to predict 3-and 5-year RFS of patients with HCC following surgical resection is shown in [fig_ref] Figure 1: Nomograms to predict probability of 3-and 5-year over survival [/fig_ref]. The model demonstrated good accuracy for predicting
## Survival analysis based on current staging system
In the current study, Barcelona Clinic Liver Cancer (BCLC) system, the seventh AJCC system, Cancer of the Liver Italian Program (CLIP) system and Hong Kong Liver Cancer (HKLC) system were used to classify all the patients. The results were shown in [fig_ref] Table 3: Variables associated with RFS according to the Cox proportional hazards model [/fig_ref]. The survival analysis stratified by BCLC system, the seventh AJCC system, CLIP system and HKLC system were present in [fig_ref] Figure 3: Kaplan-Meier survival curves of the training set [/fig_ref] and [fig_ref] Table 4: Overall survival analysis stratified by staging systems in the training set and... [/fig_ref].
## Comparsion of the nomogram and current staging system
The c-index for OS in the training set and the validation set were 0.767 (95% CI 0.742-0.792) and 0.719 (95% CI 0.671-0.767), respectively. In the training set, There were significant differences in the C-indices between the nomogram and other staging systems (0.767 vs. BCLC: 0.644; the seventh AJCC system: 0.678; CLIP: 0.692 and HKLC system; 0.688; P < 0.001 for all, respectively), and all C-indices were significantly lower than that of the nomogram. In the validation set, There were also significant differences in the C-indices between the nomogram and other staging systems (0.719 vs. BCLC: 0.609; the seventh AJCC system: 0.674; CLIP: 0.648 and HKLC system; 0.639; P < 0.001 for all, respectively), and all c-indices were significantly lower than that of the nomogram.
# Discussion
Currently, surgical resection remains the curative treatment modality for hepatocelluar carcinoma without metastasis. [bib_ref] The global burden of liver disease: the major impact of China, Wang [/bib_ref] [bib_ref] Hepatic resection associated with good survival for selected patients with intermediate and..., Zhong [/bib_ref] Hepatectomy was considered when all tumors on preoperative imaging studies could technically be resected and preserved liver function is well. Recently, many hepatobiliary institutions, including western and Oncotarget 80787 www.impactjournals.com/oncotarget eastern liver cancer centers, advocate hepatectomy to treat HCC outside the Milan criteria. [bib_ref] Hepatic resection associated with good survival for selected patients with intermediate and..., Zhong [/bib_ref] [bib_ref] Surgical treatment of hepatocellular carcinoma associated with hepatic vein tumor thrombosis, Kokudo [/bib_ref] Personal prognostic prediction is essential to guild post-operative additional treatment and postoperative counseling. The prognosis of HCC patients varies greatly according the tumor status, liver function and performance status. In clinical, for those with resectable HCCs, they almost had Child-Pugh status A even they had intermediate-advanced stage HCCs. Consistently, in this study, Child-Pugh status or liver cirrhosis did not impact the prognosis after hepatectomy. [bib_ref] Hepatic resection associated with good survival for selected patients with intermediate and..., Zhong [/bib_ref] [bib_ref] Surgical treatment of hepatocellular carcinoma associated with hepatic vein tumor thrombosis, Kokudo [/bib_ref] [bib_ref] Differences between hepatocellular carcinoma and hepatitis B virus infection in patients with..., Kim [/bib_ref] Surgery type was considered as predictor of prognosis in the Cox model. However, according to the pearson correlation analysis, surgery type and tumor size correlated closely. In clinical practice, patients with large tumor size might face that more liver tissue were be removed in order to eliminate tumor. We therefore select tumor size for analysis, namely tumor size and then construct this nomogram by combination of tumor characteristics and systematic inflammation.
Tumor characteristics in our nomogram included AFP level, tumor size, tumor number, MVI, major vascular invasion and satellite lesions. In consistent with previous study, all the variables were validated as independent risk factors in our large sample size. [bib_ref] Hepatic resection associated with good survival for selected patients with intermediate and..., Zhong [/bib_ref] [bib_ref] Importance of tumor size as a prognostic factor after partial liver resection..., Goh [/bib_ref] The cutoff value of AFP was 400 ng/ml as previous study showed. [bib_ref] Preoperative serum alpha-fetoprotein and prognosis after hepatectomy for hepatocellular carcinoma, Yang [/bib_ref] AFP > 400 ng/ml was assigned score of 10.4 points for OS prediction and 15 points for RFS prediction Tumor size was ranging from 1.2cm to 26.0cm and tumor size with 26.0 cm was assigned score of 100 points for OS prediction and RFS prediction. Tumor number was classified as one, two and more with scores of 0, 12.2 and 24.4 points, respectively. The MVI, major vascular invasion and satellite lesions obtained score of 18.0 points, 38.5 points and 13.6 points, respectively for OS prediction and 14.1 points, 53.8 points and 9.4 points, respectively for RFS prediction. In contrast to many previous studies, we included tumor size as continuous variable because various cutoff of tumor size are validated as risk factor for HCC. [bib_ref] Importance of tumor size as a prognostic factor after partial liver resection..., Goh [/bib_ref] [bib_ref] Long-term outcomes comparing surgery to embolizationablation for treatment of solitary HCC <..., Elnekave [/bib_ref] The higher the points were, the larger the tumor size was. Our previous study suggested tumor number > 3 decreased the prognosis after hepatectomy. [bib_ref] Barcelona clinic liver cancer stage B hepatocellular carcinoma: transarterial chemoembolization or hepatic..., Jianyong [/bib_ref] In our nomogram, tumor number > = 3 had the max score of about 25.0 points for OS prediction and 40 points for RFS prediction. Satellite lesion and MVI were currently acceptable risk factor associated with HCC patients' prognosis.
The relationship between systematic inflammation, such as PLR and NLR, and cancerous prognosis has been universally accepted by many investigators. Elevated NLR and PLR were identified as independent prognostic factors [bib_ref] Neutrophil to lymphocyte ratio as an indicator of the malignant behaviour of..., Okamura [/bib_ref] [bib_ref] Platelet-to-lymphocyte ratio acts as a prognostic factor for patients with advanced hepatocellular..., Li [/bib_ref] The systemic inflammatory response such as NF-κB pathway, might lead to aberrant release of proinflammatory cytokines and inflammatory mediators, promoting the tumor to proliferate and metastasize via the promotion of angiogenesis, DNA damage, and apoptosis inhibition.In the current study, we defined the mean value of PLR and NLR as cutoff value, and NLR (> 3) were demonstrated as independent risk factor. Elevated NLR got the score of appropriate 12.7 points for OS prediction and 11.7 points for RFS prediction. In our nomograms, the predictive power was enhanced when incorporating these risk predictors, with index of 0.767 for OS prediction and 0.746 for RFS prediction.
According to our nomogram, each individual could achieve a predicted 3-and 5-year OS and RFS after hepatectomy, which cannot be realized by the conventional staging systems placing patients into prognostic groups. For example, BCLC system simply stratified the patients into three groups. For the other three systems, they might not distinguish the prognosis between some stages in the training or validation set. Notably, our nomograms could pay sufficient attention to pathological information, Oncotarget 80790 www.impactjournals.com/oncotarget such as MVI and satellite lesion. [bib_ref] Proposal and validation of a new model to estimate survival for hepatocellular..., Liu [/bib_ref] [bib_ref] Nomograms for survival prediction in patients undergoing liver resection for hepatitis B..., Li [/bib_ref] [bib_ref] Prognostic Nomograms for Pre-and Postoperative Predictions of Long-Term Survival for Patients Who..., Li [/bib_ref] Moreover, in current study, we firstly incorporated the systematic inflammation into a model to predict the prognosis since this simple and effective blood marker could predict various cancerous prognosis. Consequently, our nomogram showed good calibration and discriminatory abilities with C-index value of 0.767 higher than that of previous nomogram(c-index:0.62) and four conventional staging systems [bib_ref] Predicting Individual Survival After Hepatectomy for Hepatocellular Carcinoma: a Novel Nomogram from..., Torzilli [/bib_ref].
In conclusion, our nomogram focus on tumor itself and imflammation status,which better predicted the individual prognosis after hepatectomy than conventional staging system. Nomogram is relatively easy to read with a simple graphic. When more HCC patients underwent hepatectomy, these nomograms are effective and practical.
The current study had several limitations. Firstly, while the nomogram was internally validated using bootstrapped calibration and validation set, multicenter studies are needed to externally validate the proposed nomogram. Secondly, our nomogram could only apply those with HCC after hepatectomy. Since hepatectomy provides the hope of curative therapy, it is necessary to accurately predict the prognosis of these patients and thus to guild the postoperative therapy. In the future, since more risk factors were indentified, a model with higher accuracy may be established. Thirdly, in the validation set, the sample size was relative small. The proposed nomograms needed to be validated in large sample size and prospective study.
# Materials and methods
## Patients
Patients were identified from a prospectively maintained database of HCC treated with hepatectomy at West China Hospital from July 1, 2007 through July 31, 2014. A cohort of 777 HCC patients were randomly with a 4:1 ratio allocated into the training set and the validation set via SPSS 20.0 software. The training set consisted of 618 patients and the validation set consisted of 159 patients. The diagnoses of HCC were all histologically confirmed based on analysis of paraffin-embedded tissue by experienced liver pathologists in West China Hospital. Clinicopathological variables including age, gender, hepatitis B virus surface antigen (HBsAg), liver cirrhosis, a-AFP, tumor size, number, major vascular invasion, MVI, satellite lesions, differentiation were obtained from our database. The cutoff value of AFP was defined as previous study described. [bib_ref] Preoperative serum alpha-fetoprotein and prognosis after hepatectomy for hepatocellular carcinoma, Yang [/bib_ref] The presence of MVI, visible only on microscopy, was described as determined from histopathological reports, while major vascular invasion were described as tumor invading the portal or/and hepatic vein which can be detected by imaging study or surgical findings. Two or more HCCs that were reported separately with a full description of histopathologic features were described as multiple HCCs. Nodules (< 2 cm) close to the main tumor were described as satellite lesions in the original pathology report. [bib_ref] Single Hepatocellular Carcinoma: Preoperative MR Imaging to Predict Early Recurrence after Curative..., An [/bib_ref] The types of hepatectomy including minor (1 segment), major (2-3 segments) and extended liver resection (> 3 segments) were determined according to the tumor location and size, liver functional reserve, cirrhosis, and estimated volume of future liver remnant based on CT data. All the patients were strictly stratified by BCLC system [bib_ref] Liver cancer: Approaching a personalized care, Bruix [/bib_ref] , the seventh AJCC system [bib_ref] Staging of hepatocellular carcinoma: assessment of the Japanese TNM and AJCC/UICC TNM..., Minagawa [/bib_ref] , CLIP systemand HKLC system [bib_ref] Development of Hong Kong Liver Cancer Staging System With Treatment Stratification for..., Yau [/bib_ref]. The inclusion criteria were as follows: (1) without extrahepatic metastasis; (2) HCC treated by hepatectomy;
(3) exact diagnosis of pathologically proven HCC; (4) HCC patients should have child-pugh status A, or child-pugh status B which can be improved to A. (5) liver tumor can be completely removed (R0), and for patients with liver cirrhosis, the residual liver volume should be at least 40% of total liver volume. The exclusion criteria were as follows:
(1) with other cancers; (2) with severe cardiovascular and pulmonary disease; (3) the major vascular invasion was involving in the trunk of portal vein or the inferior vena cava; (4) imaging study (liver ultrasound or abdominal contrast-enhanced computed tomography (CT) scan suggests severe liver cirrhosis or/and ascites (5) incomplete clinicopathological reports and follow-up data. Informed consent for using their data in research was obtained from all patients, and the study protocol was approved by the Ethic Committee of West China Hospital.
## Follow-up
The median follow-up for our cohort was 33.13 months in the training set and 34.80 months in the validation set. All of the patients were regularly followed at first month and every 3 months after surgery. Routing blood tests, liver function tests, AFP levels measurements, HBV-DNA levels, liver ultrasound were performed at each follow up. Once HCC recurrence including intrahepatic or/and extrahepatic recurrence was suspected, CT or/ and magnetic resonance imaging (MRI) was chosen to conform the lesions, and chest CT scan and bone scan were necessary. After HCC recurrence, Patients were evaluated at multidisciplinary team (MDT) in West China Hospital for treatment guidance based on the status of tumor and general condition. The MDT mainly comprised of experienced hepatic surgeon, radiologists, pathologists and oncologists in West China Hospital. Liver transplantation, resection, ablation, and TACE and palliative therapy could be applied. Patients were administrated anti-virus therapy, such as Entecavir (0.5 mg/day), if their HBV-DNA levels were > 1.00E + 03 copies/ml before and after surgery during follow up. RFS and OS were calculated from the date of initial treatment until the date of detection of recurrence and death or the date of last follow-up.
# Statistical analysis
Continuous variables were displayed as mean ± standard deviation and compared by Student's t test or Mann-Whitney U test (non-normal distribution data). Categorical data were shown as frequency and assessed by Fisher's exact test and two-tailed c 2 test. Survival curves stratified by various stage systems were evaluated using the Kaplan-Meier method and compared using the log-rank test.
Potential risk factors significant in univariate analysis were included multivariate analysis with forward step-wise selection process. We subsequently used the risk factors from multivarate analysis of the training set to construct a nomogram. Harrell's concordance index (c-index) ranging from 0.5 (randomness) to 1 (perfect discrimination) was used to evaluate the concordance between predicted and actual observed responses of patients. [bib_ref] Nomograms incorporating serum C-reactive protein effectively predict mortality before and after surgical..., Hsiao [/bib_ref] Being concordant means the nomogram assigned a higher probability of death (alive) to the patient who died (alive) than the one alive (death). The c-index is the probability of being concordant out of all predicted and actual patient pairs. [bib_ref] A preoperative nomogram to predict the risk of perioperative mortality following gastric..., Dhir [/bib_ref].
Calibration plots were generated to investigate the performance characteristics of our nomogram at 3 and 5 years after surgery in the training set and the validation set. Nomogram was internally validated using bootstrapping with 1000 resamples. Moreover, we estimate the predictive performance when the models are applied to new patients in the validation set. The predict power between the nomogram and other four current staging systems in the training set and the validation set were assessed using the rcorrp.cens package in Hmisc in R and were shown as the C-index. A higher C-index indicated a higher predict power for HCC. A P < 0.05 was considered statistically significant. All analyses were carried out using R version 3.3.0 with the rms packages (http://www.R-project.org) and SPSS version 20.0 (Chicago, IL, USA).
[fig] Figure 1: Nomograms to predict probability of 3-and 5-year over survival. (A) and recurrence free survival (B) for HCC treated by hepatectomy To calculate the probability of postoperative survival, we first draw a vertical line to the points scale to get the value for each factor and then sum up all the individual values. We finally draw a vertical line from the total points scale to the probability at the probability at the year 3 line to obtain 3-year survival rate and at the year 5 line to obtain 5-year survival rate. AFP: 0 = < 400 ng/ml, 1 = > 400 ng/ml; MVI: 0 = negative,1 = positive; Satellite lesions: 0 = negative,1 = positive; Tumor number: 1=one,2=two,3=more; Major vascular invasion: 0 = negative,1 = positive; NLR: 0 = < 3,1 = ≥ 3. Oncotarget 80788 www.impactjournals.com/oncotarget [/fig]
[fig] Figure 2: Good calibration of the nomogram in the training and validation set. The nomogram predicted the probabilities of postoperative 3 years survival (A), 5 years survival (B) in the training set and 3 years survival (C), 5 years survival (D) in the validation set. Oncotarget 80789 www.impactjournals.com/oncotarget [/fig]
[fig] Figure 3: Kaplan-Meier survival curves of the training set. (A) Barcelona Clinic Liver Cancer (BCLC) staging system; (B) American Joint Committee on Cancer (AJCC) seventh edition; (C) Cancer of the Liver Italian Program (CLIP) staging system; (D) Hong Kong Liver Cancer (HKLC) staging system) and the validation set ((E) Barcelona Clinic Liver Cancer (BCLC) staging system; (F) American Joint Committee on Cancer (AJCC) seventh edition; (G) Cancer of the Liver Italian Program (CLIP) staging system; (H) Hong Kong Liver Cancer (HKLC) staging system) classified by different staging systems. [/fig]
[table] Table 1: In the training set, there were 150 patients (24.3%) with age > 60 year and 541 male patients.540 patients (87.4%) had HBV infection. The average tumor size was 6.7 cm ranging from1.2 cm to 25.0 cm. There were 478 patients with solitary HCC, 82 patients (13.3%) with two tumors and 58 patients (9.4%) with more tumors. Major vascular invasion was noted in 98 patients (15.9%) and MVI in 342 (37.5%). Satellite lesions were available for 96 patients (15.5%). In validation set, the distribution of these characteristics is almost similar to the training set(Table 1). [/table]
[table] Table 2: Variables associated with OS according to the Cox proportional hazard model HR hazard ration; CI confidence interval; HBsAg hepatitis B virus antigen; AFP alpha-fetoprotein, HBV-DNA hepatitis B virus DNA; MVI microvascular invasion, TBIL total bilirubin AST aspartate aminotransferase, ALT alanine aminotransferase ; ALB albumin; TBIL total bilirubin; PLR platelet to lymphocyte ratio; NLR neutrophil to lymphocyte ratio; PT prothrombin time. [/table]
[table] Table 3: Variables associated with RFS according to the Cox proportional hazards model [/table]
[table] Table 4: Overall survival analysis stratified by staging systems in the training set and validation set BCLC Barcelona Clinic Liver Cancer staging system; AJCC American Joint Committee on Cancer seventh edition; CLIP Cancer of the Liver Italian Program staging system; HKC Hong Kong Liver Cancer staging system. www.impactjournals.com/oncotarget [/table]
|
Acute Perforated Appendicitis Complicated by Necrotizing Fasciitis and Bladder Perforation
# Introduction
Acute appendicitis is a common surgical disease that is well-studied with an overall mortality rate of 1%. However, it can lead to perforation and subsequent complications such as periappendiceal abscess or peritonitis, especially if disease recognition and management are delayed. In the case of perforation, the mortality increases nearly six-fold. Complications that follow perforation are rare. Such cases are therefore frequently treated on a case-by-case basis without definitive evidence behind each step in management. Necrotizing fasciitis is such a complication of acute appendicitis. It has an estimated mortality rate that widely varies between 6%-76%. According to our search of the literature, there have been 15 reported cases associated with acute appendicitis, with a mortality rate of 33.3% . An even rarer complication is perforation of the bladder wall, of which we found three similar cases in the literature. We report a case of perforated appendicitis causing necrotizing fasciitis of the abdominal wall and perforation of the bladder. Although our patient eventually died from her disease process, this case highlights how acute appendicitis can lead to rare complications where there are no established standards of management.
## Case presentation
An 81-year-old female with a past medical history significant for coronary artery disease, active left femoral vein deep vein thrombosis on warfarin, hypothyroidism, atrial fibrillation, and type 2 diabetes presented to the emergency department from a skilled nursing facility for evaluation of abdominal pain. The patient had been experiencing four days of abdominal pain, nausea, vomiting, and diarrhea. She noted her vomitus had a feculent odor. No blood was noted in the patient's emesis or bowel movements. The patient reported that "food goes right through me".
On physical examination, the patient was ill-appearing with a heart rate of 150/min in rapid atrial fibrillation, respiratory rate of 22/min, blood pressure of 100/60 mmHg, and oral temperature of 37.5°C. Abdominal exam was consistent with peritonitis with guarding in all four quadrants of the abdomen, which was most significant in the right lower quadrant. Of note, there was no palpable crepitus or overlying skin changes. Resuscitation was initiated with two liters of normal saline IV fluids.
Labs demonstrated leukocytosis to 19,600/μL, hemoglobin of 9.0 g/dL, and a platelet count of 370,000/μL. Her international normalized ratio (INR) was 2.6. A metabolic panel was significant for creatinine 1.56, a significant increase from baseline. Her lactic acid was elevated at 4.0. An abdominal CT scan with intravenous contrast was obtained that showed a 9.0 x 4.8 cm fluid collection and gas anterior to the bladder . The gas appeared to track upward into the anterior abdominal wall with extensive subcutaneous gas present . Also noted were thickening of the cecum and appendix and bilateral hydronephrosis and hydroureter .
## Figure 4: ct scan -coronal view -a thickened appendix (white arrow) with subcutaneous emphysema seen on both sides of the abdominal wall (white stars)
General surgery service was consulted for emergent laparotomy after anticoagulation reversal. She was given four units of fresh frozen plasma and vitamin K 10 mg IV for warfarin reversal, and ceftriaxone 1 g IV, and metronidazole 500 mg IV. separation of the peritoneum from the overlying necrotic transversalis fascia extending down into the pelvis. A dense fibrotic ring covering the base of the cecum was discovered, and the appendix had perforated. There was a large collection of dishwater-like, foul-smelling fluid that extended deep into the space of Retzius (i.e. retropubic space). This collection was consistent with periappendiceal abscess. Its cultures grew out Escherichia coli, Pseudomonas aeruginosa, and Proteus mirabilis. The dome of the bladder was noted to be necrotic and ruptured, and in direct contact with the abscess. In summary, the general surgeon performed an appendectomy, drainage of the periappendiceal abscess, repair of the bladder perforation, resection of necrotic tissue of the abdominal wall, and abdominal washout prior to closure of the abdomen. Urology service was consulted intraoperatively for assistance in the management of bladder perforation via telephone.
Postoperatively, the patient remained intubated in critical condition, requiring low-dose norepinephrine infusion. Antibiotics were tailored to the cultures, with a regimen of clindamycin, cefepime, and metronidazole. On post-operative day 13, the patient's urine output decreased significantly, and a repeat CT scan was performed, demonstrating free fluid in the abdomen likely representing urine, as well as persistent subcutaneous gas in the abdominal wall. The patient was taken back to the operating room, and it was found that she had progressive necrosis of the bladder wall with a perforation at the site of the initial repair and gross spillage of urine. There was also evidence of ongoing necrotizing fasciitis of the peritoneum and anterior abdominal wall, contributing to the continued septic shock. She underwent abdominal wall debridement, washout, and repeat repair of the bladder perforation. On post-operative day 19, the patient's septic shock did not improve and care was withdrawn, and the patient died shortly thereafter.
# Discussion
Necrotizing fasciitis is a rare and predominantly polymicrobial, gas-forming infection that aggressively spreads along the fascial layer. It can rapidly lead to sepsis and death, if untreated. Physical exam findings may include crepitus, diffuse tenderness, and rapidly spreading erythema. Early surgical debridement remains the key intervention for survival in addition to broad-spectrum antibiotics. The empiric antibiotic regimen must cover anaerobic agents, and the initial recommended treatment is piperacillin-tazobactam plus vancomycin plus clindamycin. Most frequently isolated bacteria from necrotizing fasciitis are Escherichia coli, Proteus mirabilis, Klebsiella pneumoniae, Staphylococcus aureus, and group D Streptococcus. An intraoperative culture from our patient's periappendiceal abscess revealed 2+ Proteus mirabilis, 2+ Escherichia coli, 1+ Pseudomonas aeruginosa.
Acute appendicitis has an incidence of 233/100,000 population and is most frequently seen in the 10-19 year old age group, and perforation is seen in approximately 25% of cases. Although appendicitis is a common surgical emergency, associated necrotizing soft tissue infection is extremely rare and carries a high risk of mortality. It is predominantly associated with perforation of the appendix and delay in diagnosis.
One prospective study noted that 65% of patients with perforation of the appendix presented more than 48 hours after onset of symptoms. Our patient presented four days after the onset of abdominal pain with symptom progression to feculent emesis. Acute perforated appendicitis poses a challenge for the clinician both in diagnosis and management. Clinical diagnosis of an acute abdomen may be evident by peritonitis on exam, but further imaging such as CT scan is essential in guiding specific management, especially when rare complications may be present. Our patient did not demonstrate crepitus or overlying skin changes on presentation. In addition, it is often difficult to distinguish other soft tissue infections from necrotizing fasciitis. A scoring guideline, Laboratory Risk Indicator for Necrotizing Fasciitis (LRINEC), has been suggested which is primarily based on laboratory values.
Complications from acute appendicitis vary widely, many of which are secondary to the effects of visceral perforation. They include bowel obstruction, abscess formation, uropathy, and gastrointestinal bleeding among others. Our patient had a perforation of the appendix and subsequent periappendiceal abscess formation with gas-forming bacteria that caused necrotizing fasciitis of the abdominal wall. By the day of presentation, her bladder which was directly posterior to the abscess had also perforated. This was the main factor in her poor outcome, as spillage of urine noted on her repeat laparotomy lead to continued septic shock. Of note, the bladder perforation was not detected until the initial laparotomy was under way on the day of presentation. The initial CT scan did not reveal this finding, although it did demonstrate proximity of the abscess to the bladder as well as gas that encircled the bladder. Therefore, in a patient with perforated appendicitis with abscess formation, a genitourinary diagnostic modality such as a CT cystography may help with operative management.
Consideration of rare but serious complications of common diseases must be integrated into routine clinical decision-making. The emergency medicine physician assumes the challenging task of not only providing medical management for patients with a surgical abdomen but also consulting the appropriate subspecialties for an optimal outcome. This is especially difficult but equally important when definitive imaging modalities such as the CT scan are unable to reveal every relevant pathology affecting the patient.
# Conclusions
This was a rare case of necrotizing fasciitis and bladder perforation precipitated by acute perforated appendicitis with abscess formation. Even though the mortality rate of acute appendicitis has decreased with advances in medicine, late complications are much less studied and carry a high mortality rate. Our case presented such a patient with a poor outcome. If a patient's physical exam demonstrates peritonitis, the bedside physician must have a high degree of suspicion for advanced complications of visceral perforation, including but not limited to abscess formation, necrotizing fasciitis, and perforation of periappendiceal structures such as the bladder.
# Additional information disclosures
Human subjects: Consent was obtained by all participants in this study. Dorinda Williams issued approval Not applicable. IRB Documentation This project does not meet the federal definition of human subjects "research" which Kent Hospital uses to determine when IRB review is necessary. IRB approval is not needed for this case report or the attached images. The imaging used in this case report do not identify the patient and patient (patient's durable power of attorney at time of death) consent was obtained for imaging to be used for educational purposes and for medical publications. I have spoken with the IRB coordinator whose contact information is below if needed. [email protected] IRB Coordinator Kent Hospital Warwick RI 02886 United States 401-737-7000 . Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work. |
Assessing Autonomic Function from Electrodermal Activity and Heart Rate Variability During Cold-Pressor Test and Emotional Challenge
activity measurement in humans refers to the invasive recording of the muscle sympathetic neural activity at the peroneal nerve 1,2 , non-invasive recordings of Heart Rate Variability (HRV) and Electrodermal activity (EDA) series show functional links with sympathetic and sympathovagal changes1,7,8.One way to functionally assess the ANS branches is through the spectral analysis of HRV series, whose spectrum has been divided in two main bands of interest: the high frequency (HF) band, comprising oscillations between 0.15 Hz and 0.4 Hz, and the low frequency (LF) band, between 0.04-0.15 Hz. While HF power has been proven to be a reliable metric of vagal cardiac tone, as long as the respiratory frequency is within the HF band, the LF power is known to be influenced by both branches of the ANS 3-5 . Nevertheless, although with several limitations, the LF/HF ratio has been regarded as a measure of sympatho-vagal balance 3-5 . In the literature several attempts have been done to tackle the quantification issues carried by the spectral analysis of HRV3,9,10. Prior attempts employed other quantification methods on either HRV or ECG derived series to assess the prevalence of sympathetic or parasympathetic modulations. Extraction of symbolic indices obtained from short term heart-period variability 11 and spectral analysis implementation on spontaneous beat-to-beat variability of RR and QT intervals 12 are among them. On the other hand, EDA, linked to the activity of eccrine sweat glands controlled by the sudomotor nerves, has been proven to be a reliable measure of sympathetic activation. This is because the sympathetic nervous system is responsible for controlling the sudomotor activity 13,14 , and its assessment is based on EDA spectral analysis 7,15 .Aiming at improving quantification of the sympatho-vagal balance, in this study we propose to combine EDA and HRV spectral estimates related to sympathetic and parasympathetic activity, respectively. ANS quantifiers from HRV and EDA signals have been exploited in recent studies in various application settings. Reviews of studies aiming at detecting emotional states and proposing automatic early stress recognition system through multimodal measurement of physiological signals (including HRV and EDA) are done in 16 and 17 , respectively. Spectral quantifier of skin conductance, indexing sympathetic activity and high frequency spectral power of HRV, indexing parasympathetic activity have been shown to be useful in the evaluation of effort and/or stress tested during auditory tasks with different difficulty levels 18 . Moreover, authors in 19 observed augmented sympathetic modulation in patients with schizophrenia by assessing the nonlinear coupling between HRV and skin conductance level.However, the multimodal approaches used in the aforementioned studies have considered the spectral quantifiers of HRV and EDA time series as separate sources of information reflecting ANS changes. In fact, to the best of our knowledge no previous study aimed at investigating the effect of combining the spectral measures from the two time series in ANS dynamics. Moreover, since sympathetic-parasympathetic interplay can effectively be estimated through HRV bispectral analysis 20 , here we also investigate a proper combination of EDA and HRV bispectral metrics. In fact, the functional analysis of HRV in the frequency domain does not fully characterize the dynamical properties of the ANS activity on cardiovascular control, while HRV nonlinear analysis, especially referring to higher-order spectra, can do so 3,4,20,21 . Because of the many interactions and feedback mechanisms between sympathetic and vagal activities, cardiac dynamics is deemed to the output of a complex, nonlinear system 3,21 . Consequently, a dynamical vagal activation concurrent with a sympathetic one may lead to heart rate increase 21 .In this study we exploit our recently proposed nonlinear point-process framework 20 , particularly referring to instantaneous spectral and bispectral HRV analysis. Briefly, this powerful statistical tool embeds the probabilistic generative mechanisms of heartbeat while providing model goodness-of-fit metrics 20 . The first-order moment of an inverse-Gaussian distribution characterizing a given heartbeat event is parametrized through a linear and nonlinear function of the past events, following a Wiener-Volterra representation with a Laguerre expansion of the kernels 20 .Since the hereby proposed sympatho-vagal balance indices combining EDA metrics with HRV spectral and bispectral measures are estimated in a time-varying fashion, we also investigate the low frequency dynamics to capture slow trends and high frequency dynamics which captures the fast dynamical oscillations using a wavelet analysis, thus uncovering indices of dynamical information beyond first-order statistics (the standard average along the time would be equivalent to the study of indices of slow activity exclusively).We validate the new proposed metrics on two experimental protocols: a standard stress test to evaluate cardiac autonomic function so-called cold-pressor test (CPT), and an emotional elicitation study using high-arousing video clips. Preliminary findings of this research were recently reported in 22,23 .The CPT test was selected due to the efficient release of sympathetic neurotransmitters leading to marked vasoconstriction having an indirect effect in baroreflex mediating interactions[24][25][26]. Prior clinical and experimental studies have demonstrated significant cardiovascular responses to the cold stimuli, 27 . Specifically, the significant effect of CPT on left ventricular function has been demonstrated in a recent study 28 . These studies have shown major changes in heart rate and blood pressure during the CPT phase in healthy subjects 29 , as well as in patients with pathological conditions such as hypertension 30 , depression 31 , and anxiety 32 . Nevertheless, due to methodological limitations results were controversial 29-32 .The emotional elicitation study has been chosen as a paradigmatic application exploiting reliable estimates of sympathovagal activity. Emotional processing, in fact, involves the activity of the so-called central autonomic network (CAN), comprising sympathovagal changes also driven by prefrontal cortex and amygdala functioning 33,34 . Accordingly, the study of emotional responses at the peripheral level has been widely investigated (see exemplary reviews in[35][36][37][38]. Nevertheless, reliable and specific ANS markers of emotional states in humans are still missing, possibly because of the non-satisfactory performance of state of the art about ANS metrics. Note that the quantifiers proposed in this study are intended to be included as functional set of ANS correlates of emotional states that can be retrieved noninvasively through wearable systems, in the frame of a nonlinear and non-stationary functional assessment. Methodological details, as well as experimental results and conclusions, follow below.Scientific RepoRtS |(2020) 10:5406 | https://doi.
Assessing Autonomic function from electrodermal Activity and Heart Rate Variability During cold-pressor test and emotional challenge Shadi Ghiasi 1* , Alberto Greco 1 , Riccardo Barbieri 2 , enzo pasquale Scilingo 1 & Gaetano Standard functional assessment of autonomic nervous system (AnS) activity on cardiovascular control relies on spectral analysis of heart rate variability (HRV) series. However, difficulties in obtaining a reliable measure of sympathetic activity from HRV spectra limits the exploitation of sympatho-vagal metrics. on the other hand, measures of electrodermal activity (eDA) have been demonstrated to provide a reliable quantifier of sympathetic dynamics. In this study we propose novel indices of phasic autonomic regulation mechanisms by combining HRV and eDA correlates and thoroughly investigating their time-varying dynamics. HRV and EDA series were gathered from 26 healthy subjects during a cold-pressor test and emotional stimuli. instantaneous linear and nonlinear (bispectral) estimates of vagal dynamics were obtained from HRV through inhomogeneous point-process models, and combined with a sensitive maker of sympathetic tone from eDA spectral power. A wavelet decomposition analysis was applied to estimate phasic components of the proposed sympatho-vagal indices. Results show significant statistical differences for the proposed indices between the cold-pressor elicitation and previous resting state. Furthermore, an accuracy of 73.08% was achieved for the automatic emotional valence recognition. the proposed nonlinear processing of phasic AnS markers brings novel insights on autonomic functioning that can be exploited in the field of affective computing and psychophysiology. The sympathetic nervous system (SNS) plays a fundamental role in driving cardiovascular control dynamics together with the other branch of the Autonomic Nervous System (ANS), the parasympathetic nervous system 1,2 , and includes the regulation of blood pressure, respiration, and heart rate. Many sympathetic neurons are part of the central nervous system (CNS) having origins in the thoracic and lumbar regions (T1-L2) in the spinal cord. Most of the sympathetic activity control on cardiovascular functioning is performed through noradrenergic neurons. In case of a drop in arterial pressure during situations such as anger, fear or excitement, the arterial baroreflex senses these changes in blood pressure through the baroreceptors, and the ANS acts to increase the vasoconstriction. As a result, the afferent input to the central autonomic nuclei is decreased, which in turn results in an increase in the sympathetic neural outflow. The resulting sympathetic influences in conjunction with the parasympathetic ones leads to an increase in heart rate, oxygen intake, and the blood supply to the heart and skeletal muscles. The release of epinephrine (adrenaline) and norepinephrine (noradrenaline), as well as the stimulation of sweat glands are also the results of a sympathetic nervous system elicitation. The sympathetic vasodilator nerves in the human skin results from a cholinergic co-transmission, which releases acetylcholine as a primary neurotransmitter. The acetylcholine binds to muscarinic receptors on sweat glands to activate sweating, defining a sympathetic skin response. An early quantification of the sympathetic activity has been performed through measurement of plasma or urinary norepinephrine. However, these indices may not be reliable due to their low sensitivity and non specific localization of the sympathetic response 1 . While a standard sympathetic
# Materials and methods
Subjects recruitment, experimental protocol, and acquisition set-up. Twenty-six right-handed young healthy volunteers (8 females) from the University of Pisa participated in the study. Subjects did not have any history of neurological and cardiovascular diseases, or alcoholic or smoking habits. They were asked to avoid coffee and alcohol, as well as strenuous exercise, at least 2 hours before the laboratory visit. Subjects aged between 21-32 (26 ± 3) with 60-100 kilograms of weight and 160-175 centimeters of height. All participants were screened using Patient Health Questionnaire (PHQ). This test intends to exclude the subjects with mental disorders for the experiment since the population under study is healthy control. Therefore, only participants with a PHQ score lower than 5 were included in the study. Valsalva maneuver and breathing holding was forbidden throughout the study. To ensure the hemodynamic stabilization, before the experiment participants were asked to sit in a comfortable chair, while watching a black screen. The experiment comprised two main phases: Cold-Pressor Test (CPT) and affective elicitation phase. The protocol timeline was as follows (see also:
- 4 minutes of resting state - up to 3 minutes of CPT - 3 minutes of resting state - 4 minutes and half of emotional stimuli through videos - 4 minutes of recovery.
The presentation order of CPT and emotional stimuli was randomized among subjects. Concerning the CPT, participants were asked to submerge their left hand (i.e, the non-dominant hand) up to wrist into a tank filled of ice and water (crushed ice) with the temperature of 0-4 degrees centigrade for a period of 3 minutes 43 . The choice of 3 minutes is consistent with the average pain threshold of healthy subjects. In case a subject could not tolerate this kind of elicitation, they were quickly moved to the next resting phase.
The emotional elicitation phase foresaw video-clips using a video-projector. The well-known Circumplex Model of Affect (CMA)was used for the emotional modeling through its valence and arousal dimensions. Valence refers to the pleasantness of the emotional perception, whereas arousal represents the strength of such a perception. In our study, the three categories of video-clips were associated with the following sessions: (1) high arousal and high positive valence, (2) neutral arousal/valence, (3) high arousal and high negative valence. The three categories have the same duration of 1 minute and 30 seconds and the presentation of the clips was randomized among subjects. The experiment ended by asking subjects for a complete self-evaluation of the stimuli through the visual analog scale questionnaire (VAS). Throughout the experiment, continuous electrocardiogram (ECG) and EDA signals were acquired using the BIOPAC MP35 device with a sampling rate of 500 Hz for further analyses. The dedicated modules, ECG100C Electrocardiogram Amplifier and EDA100C Electrodermal Activity Amplifier from BIOPAC inc. were used to acquire the ECG signal and the electrodermal activity, respectively. The ECG100C ECG Amplifier has a noise voltage of 0.1 V (rms) in the bandwidth 0.05-35 Hz, a CMRR of 110 dB, a 16-bit ADC, and an amplitude resolution of 0.4 mV to record the D2 lead ECG signal. Pregelled Ag/ AgCl electrodes for the ECG acquisition were placed following the Einthoven triangle configuration. EDA sensors were placed on the distal phalanx of the second and third finger of the dominant hand, imposing a DC voltage of 0.5V.
The RR series, defined as the interval between two successive QRS complexes, were derived by applying the automatic QRS complex detection algorithm on acquired ECG signals. The technical artifacts due to the errors in R-peak detection algorithm were removed by applying cubic spline interpolation method. The implementation was performed using Matlab software (R2017b version) and the Kubios HRV software. The experimental procedure was approved by the "Comitato Etico Regionale per la Sperimentazione Clinica della Regione Toscana", section "Area Vasta Nord Ovest" -Protocol n. 7803, Registry number 1072, approved on 18 Jan 2018. The recordings were carried out in agreement with the Declaration of Helsinki. Written informed consent was obtained from all subjects.
instantaneous estimates of vagal activity from inhomogeneous point-processes for heartbeat dynamics. Full details in this methodological framework are reported in 20 , referring to the recently proposed nonlinear parametric point-process model continuously characterizing the probabilistic properties of heartbeat events.
Briefly, assuming the history dependence, an inverse-Gaussian probability density function (f(t|H t , ξ(t))), parametrized in its first-order moment (i.e., the instantaneous mean μ RR (t, H t , ξ(t))) and the shape parameter
[formula] ξ > t ( ) 0 a 0 [/formula]
, characterizes each heartbeat event as a function of the history H t = (u j , RR j , RR j−1 , . . . , RR j−M+1 ), with , which are estimated through a maximum log-likelihood procedure with Newton-Raphson algorithm. Note also that this formulation is used to pre-process all heartbeat data to identify and eventually correct errors including peak mis-detection and ectopic beats.
The first-order moment is defined as a nonlinear Volterra-Wiener autoregressive model:
[formula] ∑ ∑∑ μ ξ = + + + = = = ∼ t H t gt g i t l k g i j t l k l k ( , , ( )) RR ( ) ( , ) ( ) ( , , ) ( ) ( ) (1) RR t N t i p i i q j q i j ( ) 0 0 1 0 0 2 where ∑ φ = − = − − − ∼ ∼ ∼ ( ) l t n ( ) ( ) RR RR (2) i n N t i Nt n N t n 1 ( ) ( ) ( ) 1 [/formula]
is the output of the Laguerre filters before time t and
[formula] ∑ φ α α α α = − − − − = − n k j i j ( ) (1 ) ( 1) (1 ) (3) i n i j i j ij j 2 1 2 0 [/formula]
is the i th -order discrete time orthonormal Laguerre function, with (n ≥ 0) and α the discrete-time Laguerre parameter 20 . Kolmogorov-Smirnov (KS) test and associated KS statistics are employed to calculate the model goodness-of-fit. The independence of the model-transfomed intervals are tested with the use of Auto-correlation plots. Through this framework, three levels of quantitative tools defined in the time, spectral, and bispectral domains can be obtained. Time domain estimates refer to the first-order (μ RR (t, H t , ξ(t))) and second-order (σ RR (t, H t , ξ(t))) moments of the underlying probability function f(t|H t , ξ(t)) characterizing the heartbeat series at each moment in time. Spectral quantifiers come from the linear power spectrum estimation Q(f, t, H t , ξ(t)), which depends on the linear terms g 1 (i, t) and related transformation from the Laguerre space to the Wiener-Volterra input-output terms. Once estimated, the continuous Q(f, t, H t , ξ(t)) can be integrated within the low frequency (LF = 0.05-0.15 Hz) and high frequency (HF = 0.15-0.5 Hz) bands to derive instantaneous sympatho-vagal and vagal estimates, respectively. On the bispectral domain, the bispectrum |Bis(f 1 , f 2 , t)| is defined from the linear and nonlinear input-output Wiener-Volterra terms, and can be integrated in the LF and HF bands to derive the following quantifiers: = .
[formula] ∫ ∫ = = . = . + + LL t B is f f t df df ( ) ( , , )(4) [/formula]
. . Note that the bispectral quantifier HH has been demonstrated to provide estimates of vagal autonomic outflow. All the introduced estimates are calculated with a 5 ms time resolution. More details about the formulation of the bispectrum can be found in instantaneous estimates of sympathetic activity from eDA. The EDA signal reflects changes in the skin conductance induced by the sweat gland activity, which is directly controlled by the sympathetic branch of the ANS. Therefore, it is possible to rely on EDA signal processing to quantify the sympathetic dynamics through a proper functional analysis in the frequency domain.
[formula] + + LH t B is f f t df df ( ) ( , , )(5)+ + HH t B is f f t df df ( ) ( , , )(6) [/formula]
Specifically, a normalized EDA series, as a result of a Z-score transformation, is downsampled to 50 Hz. Afterwards, a time-frequency representation of the series is obtained through short-time Fourier transform and Welch periodogram, with a Blackman window of 60s and 59s overlap. Hence, a time-varying estimate of sympathetic outflow EDA Symp(k) in k discrete time samples, within the interval (0, T), is obtained by integrating the time-frequency plane within the (0.045-0.25 Hz) band.
Previously, a convex optimization modeling approach has been proposed for the decomposition of the skin conductance measure of EDA. As the result of this decomposition, low and high frequency components of the skin conductance response (SCR) known as tonic and phasic activities, respectively, are separated. Mainly the SCR has been quantified by calculating the number of significant (above threshold) phasic driver peaks (EDA.nSCR), the sum of SCR amplitudes with respect to significant peaks (EDA.SumAmpSCR), the maximum value of phasic activity (EDA.PhasicMax). The quantification of the tonic component is done through the mean tonic activity computation (EDA.Tonic) as well. www.nature.com/scientificreports www.nature.com/scientificreports/ proposed instantaneous indices of sympatho-vagal balance and related phasic dynamics. Aiming to derive reliable sympatho-vagal balance estimators, we combined the aforementioned EDA Symp (k), taken as a consistent measure of the sympathetic tone, and point-process derived measures from HRV linked to vagal activity, namely spectral power HF and bispectral powers LL, LH, and HH. Specifically, the following sympatho-vagal measures are defined:
[formula] = = S k EDA k HF t ( ) ( ) ( ) (7) HF Symp pp t k = = S k EDA k HH t ( ) ( ) ( ) (8) HH Symp pp t k = = S k EDA k LH t ( ) ( ) ( ) (9) LH Symp pp t k = = S k EDA k LL t ( ) ( ) ( ) (10) LL Symp pp t k [/formula]
The derivation of sympatho-vagal measures is summarized in. We take a step further characterizing sympatho-vagal balance by (i) decomposing their time-varying dynamics into two components (1) the tonic activity representing a slow trend dynamic and (2) superimposed phasic component known to reflect high frequency oscillations of the time series and (ii) quantifying phasic dynamics using the median and area under the curve (AUC) metrics. Tonic and phasic decomposition of time-varying bispectral estimates LL, LH, and HH is further performed. An exemplary decomposition is shown in. Using a wavelet decomposition procedure, tonic (S ton ) and phasic (S ph ) dynamics were derived from the first level of approximation coefficients and fifth level of detail coefficients, respectively, with a Daubechie5 function as the mother wavelet Classification for automatic valence recognition. Features were grouped into six categories. The first category F 1 is from HRV measures, including instantaneous point-process estimates defined in the time and frequency domains and standard HRV time domain measures including Root Mean Square of the Successive Differences (RMSSD) and normalized mean number of times an hour in which the change in successive normal sinus (NN) intervals exceeds 50 ms (pNN50). .
[formula] F 1 ∈ [μ RR , σ RR 2 , RMSSD, [/formula]
Time-varying dynamics were averaged within a 90s time window, which corresponds to the duration of either the neutral, pleasant and unpleasant elicitation. In order to minimize the inter-subject variability, for each subject the feature set related to a negative or positive stimulus was normalized by dividing for the value computed during a neutral session. These features were used to automatically classify emotional states associated with negative and positive elicitations through a standard nonlinear Support Vector Machine (SVM) with recursive feature elimination (RFE), which was validated through a Leave-One-Subject-Out (LOSO) procedure. RFE is a wrapper feature selection method that chooses a subset of features with size r among m features (r < m) that optimizes the performance of the SVM classifier. The method is based on a backward sequential selection. Particularly, the feature that has the least impact on the SVM weight-vector norm is eliminated for each repetition. Therefore, the features are ranked and the SVM classification is repeated m times while the last ranked features are removed at each iteration. In this study, a recently developed nonlinear SVM-RFE which is an improved version of the previous RFE algorithm is implemented. This developed method tackles the issue of the correlation bias of the previous method which makes it reliable if the input feature set contains highly correlated features.
The LOSO validation scheme foresees that, at each iteration, the SVM-RFE model is trained using feature samples from N − 1 subjects (where N is the total number of subjects), and tested on the feature samples from the left-out subject. This is iterated N times. Final feature ranking was obtained by summation of the ranks from each iteration.
Classification results are reported in terms of confusion matrix associated with the feature subset giving highest accuracy. Note that diagonal values of this matrix represent sensitivity and specificity of classification, which are defined as the proportion of actual positives and negative valence elicitations that were correctly identified, respectively. The recognition was performed using MATLAB v8.4 and an additional toolbox for classification. The processing chain of the classification for valence recognition is depicted in.
# Experimental results
We first report on a comprehensive characterization in the statistical sense of the proposed features performance by comparing resting vs. CPT phases. Given the non-Gaussianity of samples, as revealed by p < 0.05 from Kolmogorov-Smirnov tests with null hypothesis of data Normality, p-values for the Rest vs. CPT comparison are from Wilcoxon non-parametric tests for paired data, with null hypothesis of equal medians between samples. These results aim at ensuring that the phasic dynamics of the proposed linear and nonlinear sympatho-vagal measures carry significant information on ANS changes. Afterwards, we exploit this knowledge in recognizing positive vs. negative valence during the emotional video elicitations by applying the aforementioned SVM-RFE algorithm. , where the feature time-varying evolution is expressed as Median ± . Among the F 1 feature set (HRV time and frequency domain features), only μ RR showed significant differences between rest and CPT for all the selected time windows. Significant differences were found also in HF pp for the 30s and 180s windows. In the F 2 feature set (HRV bispectral features), significant differences were found for LL pp at all time windows, as well as for LH pp taking estimates from 180s CPT. Remarkably, significant differences were found for the proposed features in F 3 and F 5 , combining sympathetic EDA with linear and nonlinear vagal dynamics from HRV, but S LL in the 180s time window. Concerning features from phasic dynamics in F 6 , significant p-values were found for S LL ph for the 90s and 180s time windows. However, calculating the AUC of such phasic dynamics, bispectral features showed differences at 90s, whereas the ones from nonlinear sympatho-vagal balance were all different except for S HH ph estimated at 30s. www.nature.com/scientificreports www.nature.com/scientificreports/ Valence recognition results. The recognition accuracy as a function of the number of features, ranked through the SVM-RFE procedure, is shown in, whereas the confusion matrix associated with the best classification accuracy is reported in. This was achieved using four features, namely S LH ph , LF pp , LH ph , and S HH ph , with a balanced accuracy of 73.08% resulting in 75% and 71.43% of positive predictive value and negative predictive values, respectively.
Note that such most informative features come from spectral power in the LF, as well as from the proposed phasic dynamics of bispectral sympatho-vagal features. The complete feature set used for the classification, ranked through the SVM-RFE procedure, is reported in. As a comparison the classification results using other feature sets including only the indices extracted from HRV (F 1 and F 2 ) and the standard EDA indices (F 4 ) are also reported, respectively).
# Discussion and conclusion
In this study we investigated novel ANS metrics linked to sympatho-vagal dynamics by combining EDA and HRV spectral powers, as well as combining EDA spectral power with HRV bispectral power. To this end, while sympathetic estimates were gathered from the electrodermal activity (EDA), vagal estimates were from HRV series. Features from phasic activity series of HRV bispectral power and combined EDA spectra were investigated as well. Preliminary results of this study can be found in.
More in detail, we combined features from EDA power spectra, namely EDA Symp , with instantaneous spectral and bispectral features from point-process models for heartbeat dynamics. The choice of a point process framework as a physiologically inspired probabilistic method lies in its uniqueness in parameterizing heartbeat www.nature.com/scientificreports www.nature.com/scientificreports/ dynamics continuously with any time resolution, without the need of an interpolation method. The quadratic autoregressive nonlinearities allows for the estimation of instantaneous bispectral measures and their decomposed high frequency component, which refers to the superimposed phasic activity. Time-varying information of such instantaneous phasic dynamics was quantified through central tendency (median) and AUC quantifiers. Previous studies accounted for first-and second-order moments exclusively to characterize instantaneous ANS activity 7,15,20,61-64 , therefore neglecting effective information from the superimposed phasic dynamics.
We first evaluated the proposed methodology during a paradigmatic ANS elicitation through prolonged cold-pressor test (CPT) in twenty six healthy volunteers. The significant decrease in μ RR following the CPT onset is consistent with previous studies (seeand references therein), and RMSSD and pNN50 also showed a significant decrease in specific time windows. Note that significant p-values were obtained from standard EDA measurements (F 4 ) at all time intervals from the CPT onset, and a delayed response in EDA Symp of about 30s from the CPT onset was also observed. We demonstrated that the sympatho-vagal balance can be better characterized by combining spectral and bispectral features derived from HRV with measures computed from EDA spectral analysis. The discriminative power of these measures is proven by the significant p-values associated with all time windows following the CPT onset, whereas the traditional LF/HF ratio was not able to discern CPT dynamics with respect to a previous resting state and. www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ Considering the 90s time windows from the CPT onset, while the HH did not show significant changes, the HH ph (AUC) was associated with a significant p-value. This result confirms the main hypothesis of this study for which relevant information on the ANS activity can be retrieved from the superimposed phasic behavior of HRV time-varying bispectral measures. Furthermore, the proposed new indices characterizing the cardiovascular control through EDA and HRV seem to provide a more effective indicator of the sympathovagal balance then traditional indices from HRV series only.
The statistical comparison between the two valence levels showed no significant differences in the proposed feature sets. However, such features allowed to discern between pleasant and unpleasant affective elicitation through an automatic decision support system. Indeed, features associated with a non-significant p-value may provide meaningful information in a machine learning framework. By applying the SVM-RFE algorithm, the discriminant power of the high oscillations of nonlinear bispectral estimates are confirmed with respect to measures obtained from first-order moments only. A balanced accuracy of 73.08% is achieved for a valence recognition through a proper selection of both spectral and higher-order spectral features and their integration with power spectral estimates of EDA and related phasic activations. As a comparison analysis, an HRV feature set including indices defined in the time and frequency domains (F 1 ) and related bispectral estimates (F 2 ) was fed to the classifier and an accuracy of 63.43% was obtained, with 64% and 62.96% as positive predictive value and negative predictive value, respectively. On the other hand, a 68.48% accuracy with positive and negative predictive values of 65.52% and 71.43%, respectively, was associated with the use of a standard EDA feature set F 4 (see. By comparing results in Tables 1 and 4, we may conclude that the proposed methodology could be a viable approach to integrate HRV and EDA dynamics in case of thermal and emotional elicitations tasks.
The different number of males and females in the dataset is a limitation of this study. Moreover, although the VAS phsychological test scores have proven the strong impact of stimuli in all subjects, there is no gold-standard procedure to evoke a sympathetic elicitation both at a cardiac and at a skin sympathetic outflow level. Furthermore, the emotional responses may not be consistent across the individuals, this increasing the inter-subject variability. However, it is important to remark that we did not aim to develop the most performant emotion recognition system, but rather we aimed to devise a computational method that effectively combines HRV and EDA information to asses autonomic changes during CPT and emotional processing. We performed the cold pressor test to induce changes in the autonomic control of cardiovascular dynamics different than an emotional elicitation task. Indeed, we could have performed other autonomic maneuvers including active and passive postural changes, lower body negative pressure, valsalva maneuver, handgrip test, and others. Moreover, the administration of the cold pressor test may have caused not negligible pain that might have affected the ANS estimates derived from HRV and EDA for further analyses. We are aware that specific lab conditions are required to study complex physiological phenomena related to emotional processing, and a direct translation of the proposed methodology to emotion recognition in ecological conditions may be challenging. Another limitation of this study is that in spite of the fact that the bispectral measure HH can be considered as a reliable index of cardiovascular vagal activity, the integration of LH and LL in the defined frequency bands may contain oscillations originating in the low frequency band which may not be linked to a a sympathetic activity exclusively.
Due to the asymmetric nature of the EDA signals, asymmetrically shaped Daubechies (dbN) wavelets were widely used to analyze the raw waveforms. In our study, db5 (chosen among db1 to db10) was the most appropriate choice of the mother wavelet for a better decomposition of tonic and phasic activities. However, a procedure to select a specific mother wavelet should be carried out in the future. Moreover, although previous studies found a correlation between EDAsymp and sympathetic activity 7,15 , sympathetic sudomotor nerve activity may be different from the cardiac sympathetic dynamics. Further studies are therefore needed to quantitatively investigate the functional relationship between different sympathetic markers from physiological series.
Despite the mentioned limitations, this study opens new horizons for the study of autonomic control mechanisms and the functional assessment of emotional responses of human. Future work will be directed to the evaluation of the proposed methodology on data gathered during orthostatisc stressors such as head-up tilt, handgrip, and lower body negative pressure, as well as from different patient population with cardiovascular or mental disorders, or autonomic dysfunctions.
## Data availability
The informed consent forms signed by the subjects prevent data from being publicly available. Data may be requested via email by researchers, upon reasonable request and verification of all ethical aspects, at: gaetano. [email protected].
## Received
|
The Imaging and Pathological Features of Metastatic Leiomyosarcoma in the Gallbladder
Uterine leiomyosarcoma is a rare and aggressive malignancy with poor overall prognosis. There have been few reports of metastatic leiomyosarcoma in the gallbladder. We report a case of a 41-year-old female who underwent total abdominal hysterectomy due to presumed uterine fibroids. The postoperative pathology revealed high-grade pleomorphic leiomyosarcoma, with involvement of the uterine serosal surface. She subsequently underwent exploratory laparotomy, followed by pelvic radiation and chemotherapy. Since initial management she has developed metastatic disease and has been under treatment and surveillance for 11 years. She has undergone multiple surgical procedures and numerous lines of systemic therapy for metastatic leiomyosarcoma, including cholecystectomy for a metastatic lesion in the gallbladder. There have been no previous reports of metastatic leiomyosarcoma in the gallbladder. Despite extensive metastatic disease this patient has had prolonged survival with multi-modality management.
# Introduction
Uterine leiomyosarcoma is a rare, aggressive smooth muscle malignancy, comprising 1-2% of uterine malignancies. [bib_ref] Prognostic factors in early-stage uterine sarcoma. A Gynecologic Oncology Group study, Major [/bib_ref] It is often diagnosed incidentally after hysterectomy or myomectomy for a presumed benign leiomyoma or other benign lesions. The prognosis of uterine leiomyosarcoma is very poor due to high rate of recurrence and metastasis. [bib_ref] Diagnosis, prognosis, and management of leiomyosarcoma: recognition of anatomic variants, Bathan [/bib_ref] Despite optimal local management, many patients die of distant metastatic disease. Metastatic sites often include the lungs, liver, brain and bone. [bib_ref] Uterine sarcomas: a review, Prat [/bib_ref] Other rare metastatic sites have been reported, such as breast, bile duct, bowel, pancreas, skull, adrenal gland and heart. [bib_ref] Breast metastasis from uterine leiomyosarcoma: a case report, Lin [/bib_ref] [bib_ref] Metastatic leiomyosarcoma of the intrapancreatic bile duct, Perysinakis [/bib_ref] [bib_ref] Obstructive small bowel metastasis from uterine leiomyosarcoma: a case report, Tunio [/bib_ref] [bib_ref] Isolated metastasis of uterine leiomyosarcoma to the pancreas: Report of a case..., Ozturk [/bib_ref] [bib_ref] Metastasis of uterine leiomyosarcoma to the pancreas, Iwamoto [/bib_ref] [bib_ref] Metastatic uterine leiomyosarcoma involving the pancreas diagnosed by EUS with fine-needle aspiration, Sweeney [/bib_ref] [bib_ref] Skull metastasis from uterine leiomyosarcoma: a case report, Yip [/bib_ref] [bib_ref] Uterine leiomyosarcoma metastasis to the skullcase report, Uchino [/bib_ref] [bib_ref] Unusual metastasis of uterine leiomyosarcoma to the adrenal gland with intravenous extension..., Yang [/bib_ref] However, to our knowledge there are no published reports of metastatic leiomyosarcoma of the gallbladder. Here, we present an extremely rare case of uterine leiomyosarcoma, with gallbladder metastasis along with lung, chest wall and thigh metastasis.
## Case report
A 41-year-old woman presented with vaginal bleeding and was found to have uterine fibroids on ultrasound. She underwent total abdominal hysterectomy at another center, and the pathology was reported as high-grade uterine leiomyosarcoma measuring 9 cm. She was then referred to our institution and underwent an exploratory laparotomy, extensive lysis of adhesions and left radical parametrectomy and omentectomy. She subsequently received 6 cycles of adjuvant gemcitabine and docetaxel followed by adjuvant radiation. Two years after the completion of adjuvant chemotherapy a surveillance scan detected lung metastases [fig_ref] Figure 1: Pathology of lung metastasis, showing the tumor involving the lung parenchyma [/fig_ref] and she subsequently underwent a left thoracoscopic wedge resection followed by 6 cycles of doxorubicin. Six months later a restaging computed tomography (CT) demonstrated increasing metastatic disease in the lungs and she was treated with 6 cycles of temozolamide. A CT scan following 6 cycles showed stable disease. She then underwent a right thoracoscopic wedge resection and was treated with 4 cycles of liposomal doxorubicin. A repeat CT showed stable lung nodules. Four cycles of trabectedin were then administered but she tolerated these poorly with grade 3 fatigue. She received 5 cycles of temozolamide with disease progression. She then underwent a left video-assisted thoracoscopic surgery resection and right thoracotomy wedge resection and was subsequently commenced on gemcitabine and docetaxel with bevacizumab. Following 6 cycles of gemcitabine and doceatexl she continued on single agent bevacizumab, with durable stable disease.
Three months following discontinuation of bevacizumab repeat imaging demonstrated a mass in the gall bladder, measuring 2.8×2.2×1.8 cm [fig_ref] Figure 2: Abdomen computed tomography demonstrated gallbladder lesion [/fig_ref]. The patient had no obvious symptoms. Six months later, followup CT demonstrated increased size of the mass, measuring 5.0×4.2×2.5 cm [fig_ref] Figure 2: Abdomen computed tomography demonstrated gallbladder lesion [/fig_ref]. She then underwent cholecystectomy. Pathology confirmed metastatic leiomyosarcoma in the gallbladder (6.5 cm), with tumor involving the wall of the gallbladder with extension into mucosa but not beyond the serosa and not involving perigallbladder fat [fig_ref] Figure 3: Pathology of gallbladder metastatic leiomyosarcoma [/fig_ref]. Immunochemistry showed the tumor cells to be strongly positive for smooth muscle actin [fig_ref] Figure 4: Immunohistochemistry for alpha-smooth muscle actin [/fig_ref].
Repeat imaging also demonstrated a right chest wall mass and a right posterior thigh mass. She underwent resection of all those lesions. Most recently, imaging showed multi-ple pancreatic nodules, Fine needle aspiration biopsy confirmed metastatic disease. She underwent standard pancreaticoduodenectomy. The pathology from all surgical specimens demonstrated morphological and imaging features consistent with metastatic leiomyosarcoma. As she had persistent lung metastases she commenced gemcitabine and navelbine with a partial response. She had 8 cycles of this schedule. She then went on to surveillance and most recently has been on pazopanib with stable disease. With multi-modality therapy she has lived for over 10 years with metastatic uterine leiomyosarcoma.
# Discussion
The common metastatic sites of uterine leiomyosarcoma include lung, liver, brain, and bone. [bib_ref] Uterine sarcomas: a review, Prat [/bib_ref] Other rare metastatic sites have been reported including pancreas, skull, adrenal gland and heart, bile duct, bowel, breast, orbital and gingival. [bib_ref] Breast metastasis from uterine leiomyosarcoma: a case report, Lin [/bib_ref] [bib_ref] Metastatic leiomyosarcoma of the intrapancreatic bile duct, Perysinakis [/bib_ref] [bib_ref] Obstructive small bowel metastasis from uterine leiomyosarcoma: a case report, Tunio [/bib_ref] [bib_ref] Isolated metastasis of uterine leiomyosarcoma to the pancreas: Report of a case..., Ozturk [/bib_ref] [bib_ref] Metastasis of uterine leiomyosarcoma to the pancreas, Iwamoto [/bib_ref] [bib_ref] Metastatic uterine leiomyosarcoma involving the pancreas diagnosed by EUS with fine-needle aspiration, Sweeney [/bib_ref] [bib_ref] Skull metastasis from uterine leiomyosarcoma: a case report, Yip [/bib_ref] [bib_ref] Uterine leiomyosarcoma metastasis to the skullcase report, Uchino [/bib_ref] [bib_ref] Unusual metastasis of uterine leiomyosarcoma to the adrenal gland with intravenous extension..., Yang [/bib_ref] [bib_ref] A case of orbital metastasis of uterine leiomyosarcoma with intracranial extension presenting..., Kim [/bib_ref] [bib_ref] Metastatic uterine leiomyosarcoma in the upper buccal gingiva misdiagnosed as an epulis, Cassoni [/bib_ref] To the best of our knowledge, we are unaware of any other published reports of uterine leiomyosarcoma with gall bladder metastasis.
Uterine leiomysarcoma generally disseminates via the blood stream, with the lungs being a common site of metastatic spread. However, the gallbladder does not have a rich blood supply, and it is a rare site for metastatic disease. [bib_ref] Primary leiomyosarcoma of gallbladder: a rare diagnosis, Savlania [/bib_ref] Our patient had lived for many years with metastatic leiomyosarcoma prior to the identification of a metastasis in the gall bladder. Several cases of primary leiomyosarcoma of the gallbladder have been reported although this is not a common primary site. [bib_ref] Primary leiomyosarcoma of gallbladder: a rare diagnosis, Savlania [/bib_ref] [bib_ref] Primary leiomyosarcoma of gallbladder, Park [/bib_ref] [bib_ref] Primary gallbladder sarcoma: a clinicopathologic study of 15 cases, heterogeneous sarcomas with..., Al-Daraji [/bib_ref] The prolonged survival of our patient with multimodality management, including systemic therapy, surgery and radiation, highlights the importance of managing patients with these rare tumors within a multi-disciplinary team. The median overall survival of patients with metastatic soft tissue sarcoma is in the region of 12-16 months. However, it is clear that in some patients prolonged survival can be achieved with aggressive management and access to novel treatment approaches. [bib_ref] Pazopanib, a multikinase angiogenesis inhibitor, in patients with relapsed or refractory advanced..., Sleijfer [/bib_ref] Our case also highlights the importance of unusual patterns of metastatic spread in patients with leiomyosarcoma. Physicians need to be aware of such potential metastatic sites so that appropriate management can be initiated.
# Conclusions
In conclusion, to our knowledge this is the first report of leiomyosarcoma uterine with metastatic disease in the gall bladder. This case also illustrates the importance of physicians being aware of unusual locations of metastasis and the applicability of multimodality therapy in patients with metastatic soft tissue sarcoma.
[fig] Figure 1: Pathology of lung metastasis, showing the tumor involving the lung parenchyma. A) 4×; B) 20×. [/fig]
[fig] Figure 3: Pathology of gallbladder metastatic leiomyosarcoma. A) 4×; B) 20×. [/fig]
[fig] Figure 4: Immunohistochemistry for alpha-smooth muscle actin (1A4) antibody. [/fig]
[fig] Figure 2: Abdomen computed tomography demonstrated gallbladder lesion. A) lnitial lesion size; B) lesion size after 6 months. [/fig]
|
Toll-Like Receptor 3 is Associated With the Risk of HCV Infection and HBV-Related Diseases
There are inconsistent data on the association of risk of hepatitis virus infection and hepatitis virus-related diseases with the tolllike receptor 3 (TLR3) gene.Several common polymorphism sites were targeted to assess the risk of HBV infection, HCV infection, and HBV-related diseases.Meta-analysis combining data for 3547 cases and 2797 controls from 8 studies was performed in this study. Pooled ORs were calculated to measure the risk of hepatitis virus infection and hepatitis virus-related diseases. Fixed-effects pooled ORs were calculated using the Mantel-Haenszel method.TheTLR3gene was associated with a significantly increased risk of HBV-related diseases among 1355 patients and 1130 controls ([pooled OR, [95%CI]: 1.30, [1.15-1.48] for dominant; 1.77, [1.35-2.31] for recessive; 1.28 [1.16-1.41] for allele frequency). Subgroup analyses by a polymorphism site indicated an increased risk of HCV infection in relation to the TT/CT genotypes of rs3775291 (1.50 [1.11-2.01]), and a decreased risk ascribed to the T allele (0.20 [0.16-0.25]). We also noted an association between rs3775291 and significantly increased risk of HBV-related diseases (2.23 [1.55-3.21]). No significant inter-study heterogeneity or publication bias was detected in the analyses.These data suggest a likely effect on the risk to infect HCV and develop HBV-related diseases for the TLR3 gene. Large-scale studies with racially diverse populations are required to validate these findings.(Medicine 95(21):e2302)Abbreviations: CI = confidence intervals, dsRNA = doublestranded RNA, HBV = hepatitis B virus, HCC = hepatocellular carcinoma, HCV = hepatitis C virus, HWE = Hardy-Weinberg equilibrium, OR = odds ratios, SNP = single-nucleotide polymorphisms, TLR3 = toll-like receptor 3.Editor: Raouf HAJJI.
# Introduction
H epatitis B virus (HBV) and hepatitis C virus (HCV) are 2 ever-increasing health problems around the globe. [bib_ref] The interferon signaling pathway genes as biomarkers of hepatitis C virus disease..., Helbig [/bib_ref] More than 2,000,000,000 and 210,000,000 people worldwide have been affected by HBV and HCV, respectively. [bib_ref] Global epidemiology of hepatitis C virus infection, Shepard [/bib_ref] The 2 types of virus are considered to be major prevalent infectious agents and leading causes of severe liver diseases such as chronic hepatitis, liver cirrhosis, and hepatocellular carcinoma (HCC). A long list of conventional risk factors for HBV and HCV acquisition have previously been identified. [bib_ref] Epidemiology of hepatitis C in Europe, Negro [/bib_ref] In addition, genetic variations, typically single-nucleotide polymorphisms (SNP), in innate immunity receptors might be susceptibility factors for hepatitis virus infection. [bib_ref] Genetic variants of innate immune receptors and infections after liver transplantation, Sanclemente [/bib_ref] The innate immune system is a very important mechanism in defense against pathogenic intrusion. Delayed recognition and detection of the presence of infecting pathogens lead to organ dysfunction, inappropriate systemic responses, devastating tissue damage, life-threatening infections, and even death. [bib_ref] Single nucleotide polymorphisms of Toll-like receptors and susceptibility to infectious diseases, Skevaki [/bib_ref] Responses of the innate immune system are inducible and can be activated by pattern recognition receptors, which stimulate conserved host defense signaling pathways that regulate the expression of many immune response genes. [bib_ref] Innate immune recognition: mechanisms and pathways, Medzhitov [/bib_ref] Toll-like receptor 3 (TLR3) is a key pattern recognition receptor of the innate immune system. The TLR3 gene triggers innate immune responses, enhances the production of cytokines essential for activation of innate immunity, and recognizes pathogen-associated molecular patterns expressed on infectious micro-organisms.
SNPs occur in >1% of the general population. They could induce amino acid conversions and thereby modify the promoter activities. [bib_ref] Genetic variants of innate immune receptors and infections after liver transplantation, Sanclemente [/bib_ref] [bib_ref] Promoter region polymorphism & expression profile of toll like receptor-3 (TLR-3) gene..., Medhi [/bib_ref] Earlier research suggests significantly higher expression levels in chronic HCV infection patients with TLR3 genotypes compared to their healthy counterparts. [bib_ref] Promoter region polymorphism & expression profile of toll like receptor-3 (TLR-3) gene..., Medhi [/bib_ref] Recent research also demonstrates evidence of a higher infection rate in individuals who harbor SNPs of the TLR3 gene that controls efficacy of innate immunity. [bib_ref] Toll-like receptors and human disease: lessons from single nucleotide polymorphisms, Lin [/bib_ref] Therefore, SNPs in TLR3 may increase the risk of hepatitis virus infection and hepatitis virus-related diseases.
Multiple SNPs positioned in the TLR3 gene, especially those under investigation (rs1879026, rs3775296, rs3775291, rs5743305), have been targeted to assess the risk of HBV-, HCV infection, and related diseases. [bib_ref] Toll-like receptor 3 polymorphism and its association with hepatitis B virus infection..., Al-Qahtani [/bib_ref] [bib_ref] Interaction of TLR-IFN and HLA polymorphisms on susceptibility of chronic HBV infection..., He [/bib_ref] [bib_ref] Genetic polymorphisms in Toll-like receptor 3 gene are associated with the risk..., Huang [/bib_ref] [bib_ref] Association between a functional polymorphism in Toll-like receptor 3 and chronic hepatitis..., Lee [/bib_ref] [bib_ref] Toll-like receptor 3 genetic variants and susceptibility to hepatocellular carcinoma and HBV-related..., Li [/bib_ref] [bib_ref] Impaired toll-like receptor 3-mediated immune responses from macrophages of patients chronically infected..., Qian [/bib_ref] [bib_ref] Association of Toll-like receptor 3 polymorphisms with chronic hepatitis B and hepatitis..., Rong [/bib_ref] [bib_ref] Toll-like receptor 3 gene polymorphisms are not associated with the risk of..., De Sá [/bib_ref] But the results are inconsistent. The objective of this study was to examine the associations between TLR3 and risk of HBV infection, HCV infection, and HBV-related diseases by means of meta-analysis.
# Methods
## Literature search
We performed a systematic literature search in PubMed, Scopus, and Embase databases, without language restrictions. Search terms included toll-like receptor 3, TLR3, CD283, hepatitis, HBV infection, HCV infection, polymorphism, polymorphisms, variants, and variations. These words were used in combination or in isolation. Two of the authors singled out all publications addressing the association of interest and examined their reference lists. They also checked the references cited in related review articles to obtain additional data. The study was approved by the institutional review board. All patients provided written informed consent.
## Selection of studies
Selection of studies eligible for inclusion was based on the following criteria defined before the commencement of publication search. Whenever 2 or more studies were overlapped in terms of subjects, we retained the most complete and informative study with a larger sample size.
## Inclusion criteria
Included patients with HBV-infection, HCV-infection or HBV-related disease, and non-infected or healthy controls; Examined the association of SNPs in the TLR3 gene with risk of HBV infection, HCV infection, or HBV-related disease; Clearly reported the genotype frequencies of each polymorphism. Genotype distribution in controls must be in accordance with Hardy-Weinberg equilibrium (HWE).
## Exclusion criteria
Editorials, reviews, comments, abstracts with insufficient data, and conference proceedings. Full-length articles with inaccessible genetic data. Only patients were included. Studies that deviated from HWE.
## Data extraction
Data for each study were separately extracted by 2 authors using a standard form. Disagreements were resolved by consensus including a third author. Information on the following characteristics was recorded: first author, year of publication, polymorphisms studied, infection type, disease type, number of genotypes, total cases and controls, location in which the study was conducted, ethnicity of study subjects, proportion of males, and study setting.
# Statistical analysis
Pooled odds ratios (ORs) were estimated to assess the relationship between TLR3 SNPs and risk of HBV-infection, HCV-infection, and HBV-related diseases. The ORs were calculated by comparing the mutated homozygote/heterozygote genotypes (22/12) with the wide-type genotype (11), then 22 with 12/11 in a recessive model, and 2 with 1 in an allele frequency model. Heterogeneity assumption was measured by the chi-square-based Q test, [bib_ref] Quantitative synthesis in systematic reviews, Lau [/bib_ref] with P value <0.05 considered significant. The summary ORs with the corresponding 95% confidence intervals (95%CI) were calculated with the fixedeffects or the random-effects model. Random-effects summary ORs were calculated using the DerSimonian-Laird method when significant heterogeneity presented. [bib_ref] Meta-analysis in clinical trials, Dersimonian [/bib_ref] The Mantel-Haenszel method was utilized to calculate fixed-effects summary ORs in case of the absence of heterogeneity. [bib_ref] Statistical aspects of the analysis of data from retrospective studies of disease, Mantel [/bib_ref] HWE was checked by the chi-square test in the controls. Publication bias was determined by funnel plot and Egger's test, 20 a type of linear regression methodology widely used to measure funnel plot asymmetry. Sensitivity analyses were performed to examine the validity of meta-analysis results. Statistical analyses were done by the STATA software (package v.12.0). P < 0.05 was considered statistically significant.
# Results
## Studies included in the meta-analysis
A total of 231 citations were identified through databases. Around 158 citations were excluded after title review. We then reviewed 73 abstracts. Among these, 59 articles not associated with TLR3 SNPs, hepatitis virus infection or hepatitis virusrelated diseases were excluded. Of the14 articles whose eligibility could not be determined by title and abstract review, 6 were discarded after full-text review. The specific reasons for study exclusion and inclusion are described in [fig_ref] FIGURE 1: The low diagram of included/excluded studies [/fig_ref]. Thus, 8 articles were included in the final analysis, [bib_ref] Toll-like receptor 3 polymorphism and its association with hepatitis B virus infection..., Al-Qahtani [/bib_ref] [bib_ref] Interaction of TLR-IFN and HLA polymorphisms on susceptibility of chronic HBV infection..., He [/bib_ref] [bib_ref] Genetic polymorphisms in Toll-like receptor 3 gene are associated with the risk..., Huang [/bib_ref] [bib_ref] Association between a functional polymorphism in Toll-like receptor 3 and chronic hepatitis..., Lee [/bib_ref] [bib_ref] Toll-like receptor 3 genetic variants and susceptibility to hepatocellular carcinoma and HBV-related..., Li [/bib_ref] [bib_ref] Impaired toll-like receptor 3-mediated immune responses from macrophages of patients chronically infected..., Qian [/bib_ref] [bib_ref] Association of Toll-like receptor 3 polymorphisms with chronic hepatitis B and hepatitis..., Rong [/bib_ref] [bib_ref] Toll-like receptor 3 gene polymorphisms are not associated with the risk of..., De Sá [/bib_ref] providing 3547 cases and 2797 controls. Asian subjects were used in most studies (62.5%, n ¼ 5). In addition, 2 studies employed North Americans and 1 enrolled South Americans. Regarding the setting of study, 5 included healthy individuals, 2 used individuals with previous HBV clearance, and 1 included non-HCV-infected liver recipients [fig_ref] TABLE 1: Primary Characteristics of Studies Incorporated in Meta-Analysis [/fig_ref].
## Association of tlr3 polymorphisms with hbv, hcv infection
There was no association observed between the TLR3 polymorphisms and HBV infection [fig_ref] TABLE 2: Impact of TLR3 on Risk of Hepatitis Virus Infection and Hepatitis Virus-Related... [/fig_ref]. .
In the subgroup analysis according to polymorphism site, there was a 1.
# Sensitivity analysis
Sensitivity analyses by meta-analysis in the absence of 1 single study were conducted to examine the stability of the combined risk estimates. Statistical significance of the summary ORs was not substantially modified (data not shown). Therefore, our results are stable.
## Publication bias
Funnel plots were created by the standard error plotting against the OR for each study. No evidence of asymmetry was indicated in the funnel plots (P > 0.05). Egger's linear regression test provided data that supported the absence of publication bias in the meta-analysis (P > 0.05).
# Discussion
As far as we are aware, this is the first study by use of metaanalysis to examine the relationship between polymorphisms in the TLR3 gene and risk of hepatitis virus infection and hepatitis virus-related diseases. Overall, we found a significantly increased risk of HBV-related diseases in relation to the polymorphisms tested. The association remained highly significant when the data were restricted to rs3775291. A strong association was also observed in the analyses of rs3775291 and HCV infection. But it is interesting that the TT/CT genotypes were associated with increased risk of HCV infection and the T allele protected against the virus infection. There was no evidence of high possibility of publication bias or significant interstudy heterogeneity.
The strong association observed in our analysis is consistent with the biological properties of TLR3. TLR3 is a known pattern-recognizing receptor with an important part in immune signaling and innate immune recognition of intruding microbes, stimulating adequate immune responses. A growing body of literature has suggested that SNPs in the TLR genes attenuate the ability of some individuals to respond appropriately to TLR ligands, and this attenuation may lead to increased occurrence of infections and infectious diseases. [bib_ref] Single nucleotide polymorphisms of Toll-like receptors and susceptibility to infectious disease, Schröder [/bib_ref] Results of a recent study support the previous finding. Ranjith-Kumar et al suggested that several polymorphisms that alter TLR3 amino acids initiate resultant changes in the protein and that they might downregulate the gene expression and lower the activities of TLR3 required for proper signaling. [bib_ref] Effects of single nucleotide polymorphisms on Toll-like receptor 3 activity and expression..., Ranjith-Kumar [/bib_ref] Reduced activity of TLR3 results in failure to recognize invading microorganisms and insufficient immune responses, thus increasing the likelihood of infections and infectious diseases. All data point to the importance of TLR3 SNPs in the development of infections and infection-induced diseases.
The association between rs3775291 polymorphism in the coding region of the TLR3 gene and risk of HCV infection and HBV-related diseases is also biologically plausible. This extensively studied polymorphism results in amino acid substitution. Mutations at this site may destroy the structure and impair the function of the protein due to the highest purifying selection pressure related to the C allele. [bib_ref] Effects of single nucleotide polymorphisms on Toll-like receptor 3 activity and expression..., Ranjith-Kumar [/bib_ref] The amino acid leucine at rs3775291 is adjacent to an asparagine whose glycan moiety binds double-stranded RNA (dsRNA), 23 a replication intermediate of many viruses. As mutations in the neighboring region (Asp413) result in significantly reduced signaling activity of TLR-3, rs3775291 may prevent glycan moiety of Asp413 from binding to dsRNA or influence its glycosylation, which seems to be able to explain the reduced signaling activity induced by rs3775291. [bib_ref] A role for Toll-like receptor 3 variants in host susceptibility to enteroviral..., Gorbea [/bib_ref] [bib_ref] Structural and functional analyses of the human Toll-like receptor 3. Role of..., Sun [/bib_ref] Again, reduction in signaling activity of TLR-3 causes higher incidence of infections and related diseases. Therefore, it is likely that the rs3775291 polymorphism is associated with the risk of HCV infection and HBV-related diseases. Our findings are also in concordance with a recent observation that HCV per se may induce downregulation of TLR3 expression on innate immune cells. [bib_ref] Expression of Toll like receptors 3 & 7 in peripheral blood from..., Mohammed [/bib_ref] Several investigations have examined the association of TLR3 polymorphisms with hepatitis virus acquisition and hepatitis virus-related diseases. Al-Qahtani et al included 707 patients infected with HBV and 600 uninfected controls of Arabian descent, demonstrating a significant effect of genetic variations in the TLR3 gene on the outcome of HBV infection. [bib_ref] Toll-like receptor 3 polymorphism and its association with hepatitis B virus infection..., Al-Qahtani [/bib_ref] The positive finding was confirmed in a replication study among more than 1400 subjects of Han Chinese descent. [bib_ref] Interaction of TLR-IFN and HLA polymorphisms on susceptibility of chronic HBV infection..., He [/bib_ref] In a most recent analysis, Huang et al likewise suggested a significant association between TLR3 SNPs and HBV-induced diseases. [bib_ref] Genetic polymorphisms in Toll-like receptor 3 gene are associated with the risk..., Huang [/bib_ref] Unlike the aforementioned analyses, Sa et al showed the absence of an association between TLR3 polymorphisms and susceptibility to HBV and HCV infection in 109 cases and 299 healthy controls of South American descent. [bib_ref] Toll-like receptor 3 gene polymorphisms are not associated with the risk of..., De Sá [/bib_ref] [bib_ref] Chronic hepatitis B in children and adolescents: epidemiology and management, Abdel-Hady [/bib_ref] With respect to HCV-related infection and related diseases, the prevalence is relatively lower in economically developed countries including Africa, the West Pacific, Eastern Mediterranean, and South-East Asia, than in less developed . Meta-analysis for association between TLR3 polymorphisms and the risk of HBV-related diseases by the fixed-effects model. countries including northern and western Europe, North America, and Australia (1-2% vs 5-10%). [bib_ref] Epidemiology of chronic hepatitis C virus infection in sub-Saharan Africa, Madhava [/bib_ref] [bib_ref] Epidemiology and natural history of hepatitis C virus infection, Lee [/bib_ref] Based on these data, we infer that there may be ethnic differences in the influence of TLR3 polymorphisms.
Some limitations should be noticed in interpreting the results. First, since most of the polymorphisms being investigated were evaluated in a low number of studies, we were not able to detect their effects individually, possibly resulting in overgeneralized results. Second, we observed significant effects when all published data were combined. However, the impact of TLR3 polymorphisms may be ethnicity-specific. It is possible that the genetic polymorphisms with a significant role in one ethnic group may not have equal impact on another. Hence analyses restricted to certain specific ethnic groups are necessary. Third, despite the potential confounding effect conferred by age, gender, lifestyle, and environmental risk factors, we were not allowed to evaluate these effects in our study as a result of data inadequacy.
In summary, polymorphisms of the TLR3 gene may be potential risk factors for HCV infection and HBV-related diseases. Further analyses incorporating a larger number of samples and ethnically different populations are warranted to determine the association between TLR3 and risk of hepatitis virus infection and hepatitis virus-related diseases.
[fig] FIGURE 1: The low diagram of included/excluded studies.Subgroup analyses by polymorphism site indicated a significant increase in the risk of HCV infection associated with the TTAssociation of TLR3 Polymorphisms With HBV-Related DiseasesUsing the Mantel-Haenszel method, the polymorphisms in the TLR3 gene were associated with a significantly increased risk of HBV-related diseases. The pooled OR was greatest in the recessive model(1.77 [1.35-2.31]), and slightly lower in dominant model (1.30 [1.15-1.48]) and allele frequency model (1.28 [1.16-1.41]). The forest plot is shown in [/fig]
[fig] FIGURE 2: Meta-analysis for association between rs3775291 and the risk of HCV infection by the fixed-effects model. HCV ¼ hepatitis C virus. [/fig]
[table] TABLE 1: Primary Characteristics of Studies Incorporated in Meta-Analysis [/table]
[table] TABLE 2: Impact of TLR3 on Risk of Hepatitis Virus Infection and Hepatitis Virus-Related Diseases Using the Fixed-Effects Model HBV ¼ hepatitis B virus, HCV ¼ hepatitis C virus, OR ¼ odds ratios, SNP ¼ single nucleotide polymorphism.&& [/table]
|
Defining Molecular Sensors to Assess Long-Term Effects of Pesticides on Carcinogenesis
The abundance of dioxins and dioxin-like pollutants has massively increased in the environment due to human activity. These chemicals are particularly persistent and accumulate in the food chain, which raises major concerns regarding long-term exposure to human health. Most dioxin-like pollutants activate the aryl hydrocarbon receptor (AhR) transcription factor, which regulates xenobiotic metabolism enzymes that belong to the cytochrome P450 1A family (that includes CYP1A1 and CYP1B1). Importantly, a crosstalk exists between estrogen receptor α (ERα) and AhR. More specifically, ERα represses the expression of the CYP1A1 gene, which encodes an enzyme that converts 17β-estradiol into 2-hydroxyestradiol. However, (ERα) does not repress the CYP1B1 gene, which encodes an enzyme that converts 17β-estradiol into 4-hydroxyestradiol, one of the most genotoxic estrogen metabolites. In this review, we discuss how chronic exposure to xenobiotic chemicals, such as pesticides, might affect the expression of genes regulated by the AhR-ERα crosstalk. Here, we focus on recent advances in the understanding of molecular mechanisms that mediate this crosstalk repression, and particularly on how ERα represses the AhR target gene CYP1A1, and could subsequently promote breast cancer. Finally, we propose that genes implicated in this crosstalk could constitute important biomarkers to assess long-term effects of pesticides on human health.
# Introduction
The aryl hydrocarbon receptor (AhR) is a ligand-activated transcription factor that belongs to the basic helix-loop-helix (bHLH)/Per-Arnt-Sim (PAS) family [bib_ref] The aryl hydrocarbon receptor complex, Hankinson [/bib_ref]. It is maintained inactive in the cytoplasm in a complex with the Hsp90-XAP2-p23 chaperones [bib_ref] Pongratz, I. The HSP90 chaperone complex regulates intracellular localization of the dioxin..., Kazlauskas [/bib_ref]. Activation of AhR occurs via direct binding of its ligands, where it translocates into the nucleus to form a complex with the bHLH/PAS Arnt protein. This heterodimer complex then binds regulatory consensus regions termed xenobiotic response elements (XREs) located at proximity to its target genes, which include the phase I detoxifying monooxygenases CYP1A1 and CYP1B1 [bib_ref] P450 genes: Structure, evolution, and regulation, Nebert [/bib_ref]. The best-characterized AhR ligands commonly fall into the following classes: halogenated aromatic hydrocarbons (HAHs), such as 2,3,7,8-tetra chlorodibenzo-p-dioxin (TCDD), polycyclic aromatic hydrocarbons (PAHs) and polychlorinated biphenyls (PCBs) [bib_ref] Chlorinated biphenyl induction of aryl hydrocarbon hydroxylase activity: A study of the..., Poland [/bib_ref]. All these chemicals are human by-products that emanate from human-based activities. For example, high levels of PAHs are found in barbecue cooked meat, and are becoming a major concern for human health considering their persistence in the environment.
AhR is best known for its role as a mediator of toxicity during environmental pollutant metabolism. However, recent studies show that AhR plays an important role in normal physiology and development as well [bib_ref] The aryl hydrocarbon receptor in immunity, Esser [/bib_ref] [bib_ref] Dioxins, the aryl hydrocarbon receptor and the central regulation of energy balance, Linden [/bib_ref] [bib_ref] The aryl hydrocarbon receptor: A perspective on potential roles in the immune..., Stevens [/bib_ref] [bib_ref] The aryl hydrocarbon receptor sans xenobiotics: Endogenous function in genetic model systems, Mcmillan [/bib_ref] [bib_ref] Characterization of a murine Ahr null allele: Involvement of the Ah receptor..., Schmidt [/bib_ref]. There has also been various studies demonstrating interactions and crosstalk between AhR and various intracellular signaling pathways, including NF-κB [bib_ref] Ah receptor and NF-κB interactions, a potential mechanism for dioxin toxicity, Tian [/bib_ref] [bib_ref] The RelA NF-kappaB subunit and the aryl hydrocarbon receptor (AhR) cooperate to..., Kim [/bib_ref] , Nrf2 [bib_ref] Cross-talk between transcription factors AhR and Nrf2: Lessons for cancer chemoprevention from..., Hayes [/bib_ref] , Rb/E2F [bib_ref] Aromatic hydrocarbon receptor interaction with the retinoblastoma protein potentiates repression of E2F-dependent..., Puga [/bib_ref] , and Sp1 [bib_ref] Regulation of constitutive gene expression through interactions of Sp1 protein with the..., Wang [/bib_ref] , as well as other transcription factors belonging to the nuclear receptor family, such as the androgen receptor [bib_ref] Dioxin receptor is a ligand-dependent E3 ubiquitin ligase, Ohtake [/bib_ref] , and the estrogen receptors (ERα and ERβ ) [bib_ref] Dioxin receptor is a ligand-dependent E3 ubiquitin ligase, Ohtake [/bib_ref] [bib_ref] ERα-AHR-ARNT protein-protein interactions mediate estradiol-dependent transrepression of dioxin-inducible gene transcription, Beischlag [/bib_ref] [bib_ref] Aryl hydrocarbon receptor-mediated transcription: Ligand-dependent recruitment of estrogen receptor alpha to 2,3,7,8-tetrachloro..., Matthews [/bib_ref] [bib_ref] Modulation of oestrogen receptor signalling by association with the activated dioxin receptor, Ohtake [/bib_ref] [bib_ref] The aryl hydrocarbon receptor mediates degradation of estrogen receptor alpha through activation..., Wormke [/bib_ref]. The crosstalk between the signaling pathways of AhR and ERα receptors will be the main focus of this review, especially in the context of breast cancer development.
AhR is proposed to have a role in breast development in utero, during pregnancy and also as previously mentioned, in breast cancer initiation. The mammary gland is composed of different cell types and structures that are subject to change during various and specific life stages, i.e., puberty and pregnancy. Consequently, exposure to AhR agonists could have different outcomes making it challenging to show direct effects of pollutants on breast cancer incidence. Several studies reported higher AhR expression in more malignant breast cancer cell lines as well as a positive correlation with breast tumors aggressiveness [bib_ref] The aryl hydrocarbon receptor: A target for breast cancer therapy, Powell [/bib_ref] [bib_ref] Constitutive regulation of CYP1B1 by the aryl hydrocarbon receptor (AhR) in pre-malignant..., Yang [/bib_ref]. However, another study showed that 72.7% of benign mammary tissues have nuclear AhR immunohistochemistry staining [bib_ref] A novel role of the aryl hydrocarbon receptor (AhR) in centrosome amplification..., Korzeniewski [/bib_ref]. The two main target genes of AhR in breast tissues are CYP1A1 and CYP1B1. CYP1A1 mRNA and protein levels are low in normal and breast tumor tissues. However, higher CYP1B1 mRNA and protein levels are observed in breast tumors as compared to normal tissues.
## Ligands and agonists of aryl hydrocarbon receptor (ahr)
The best-characterized high-affinity AhR ligands are hydrophobic molecules that bear aromatic carbon rings. However, there is more and more evidence suggesting that there is a growing spectrum of structurally diverse chemicals that are capable of binding and/or activating the AhR signaling pathway. In this section, well will briefly discuss known endogenous and exogenous ligands of AhR, with an emphasis on pesticides.
## Endogenous ligands
Human exposure to toxic environmental chemicals has greatly changed in the last 200 years, especially in societies struck by the industrial revolution. On the one hand, this suggests that AhR has probably been the target of recent evolutionary pressure in various species. One example recently published by Mark Hahn's laboratory [bib_ref] Mechanistic basis of resistance to PCBs in Atlantic tomcod from the Hudson..., Wirgin [/bib_ref] showed that Atlantic tomcod from the Hudson River exposed to high level of PCB, released by General Electric facilities, exhibited a variant of the AHR2 gene, which is absent in nearly all tomcod from elsewhere. This variant possesses less affinity for TCDD and others AhR ligands, and consequently has less transcriptional activity and toxicity. On the other hand, it is reasonable to assume that AhR response to these man-made chemicals is a sign of biochemical versatility of the receptor to bind a wide range of molecules (see below). This, in turn, may suggest that AhR could also respond to various endogenous ligands, i.e., produced by the cells themselves. Studies showing that the AhR signaling pathway is active in the absence of exogenous ligands have reinforced this view, and lead scientists to suspect a role for AhR in physiological functions of cells. Consequently, intensification of research to find endogenous AhR ligands has sparked over the last few years. The following chemicals, which are endogenously synthesized by human tissues, are known activators of AhR: indigoids, equilenin, heme metabolites, arachidonic acid metabolites, eicosanoids, and tryptophan derivatives [bib_ref] Endogenous and exogenous ligands of aryl hydrocarbon receptor: Current state of art, Stejskalova [/bib_ref] [bib_ref] Activation of the aryl hydrocarbon receptor by structurally diverse exogenous and endogenous..., Denison [/bib_ref] [bib_ref] The search for endogenous activators of the aryl hydrocarbon receptor, Nguyen [/bib_ref]. L-Kynurenine, a tryptophan catabolite, was shown to activate AhR and to be constitutively produced by tumor cells, and astonishingly promote their survival and escape from the immune system [bib_ref] An endogenous tumour-promoting ligand of the human aryl hydrocarbon receptor, Opitz [/bib_ref]. The later observations provide new evidence that an endogenous ligand of AhR promotes carcinogenesis, thereby conferring oncogenic properties to AhR.
## Exogenous ligands
## "classical" synthetic ahr ligands
As mentioned earlier, AhR is activated by a large group of environmental pollutants composed of HAHs (including TCDD), PAHs (including benzo[α]pyrene), and PCBs [bib_ref] Chlorinated biphenyl induction of aryl hydrocarbon hydroxylase activity: A study of the..., Poland [/bib_ref]. The majority of these chemicals are formed and released as by-products of human activities, mostly emanating from industrial and combustion processes [bib_ref] Stavljenic Rukavina, A. Dioxins and human toxicity, Marinkovic [/bib_ref].
## Natural/dietary compounds
One of the most obvious potential sources of naturally occurring AhR ligands probably comes from our diet. The non-toxic indole-3-carbinol (I3C) and its derivatives, including 3,3'-diindolylmethane (DIM), have been gathering great attention lately for their anticancer properties. However, there are other natural compounds that have been reported to activate the AhR signaling pathway, including flavonoids, carotenoids, curcumin, and others, which are reviewed in Nguyen et al. and Stejskalova et al. [bib_ref] Endogenous and exogenous ligands of aryl hydrocarbon receptor: Current state of art, Stejskalova [/bib_ref] [bib_ref] The search for endogenous activators of the aryl hydrocarbon receptor, Nguyen [/bib_ref].
## Pesticides
In depth analysis of a wide range of pesticides currently used in agriculture revealed that many of them possess a structure very similar to well-known AhR ligands, such as TCDD. Considering that exposure to pesticides may lead to various human diseases, including cancer, the implication of AhR in this process should be carefully assessed. It has been documented for quite some time now that pesticides possess endocrine disrupting properties [bib_ref] Developmental effects of endocrine-disrupting chemicals in wildlife and humans, Colborn [/bib_ref]. Several currently used pesticides have been reported to have estrogenic activity [bib_ref] Estrogenic action of DDT and its analogs, Welch [/bib_ref] [bib_ref] Studies on the in vivo and in vitro estrogenic activities of methoxychlor..., Bulger [/bib_ref] [bib_ref] Methoxychlor as a model for environmental estrogens, Cummings [/bib_ref] [bib_ref] Effects of the β-isomer of hexachlorocyclohexane on estrogen-sensitive human mammary tumor cells, Coosen [/bib_ref] [bib_ref] The E-SCREEN assay as a tool to identify estrogens: An update on..., Soto [/bib_ref] [bib_ref] Estrogenic and antiprogestagenic activities of pyrethroid insecticides, Garey [/bib_ref] [bib_ref] Effects of currently used pesticides in assays for estrogenicity, andrenogenicity, and aromatase..., Anderson [/bib_ref]. Endocrine disruption of the estrogen receptor-signaling pathway can be direct (i.e., chemicals bind the estrogen receptor and modulate its activity) or indirect (i.e., chemicals affect another pathway such as the AhR case, which then modulates ERα activity). More recently, in vitro reporter gene assays to screen 200 pesticides for estrogen and androgen activities. They also tested the AhR agonistic activity for these 200 pesticides. Out of those, three herbicides-propanil, linuron and diuron-showed potent AhR agonistic activity and only two-chlorpyrifos and isoxathion-showed both AhR and ERα activities [bib_ref] In vitro screening for aryl hydrocarbon receptor agonistic activity in 200 pesticides..., Takeuchi [/bib_ref] [bib_ref] Screening for estrogen and androgen receptor activities in 200 pesticides by in..., Kojima [/bib_ref].
Several studies were able to measure pesticide concentrations in breast cancer biopsies (adipose tissues and tumor sections). As such, Cassidy and collaborators showed that heptachlor epoxide induces nitric oxide production in the breast tissues, which may contribute to tumor initiation by increasing DNA damages in the cells [bib_ref] The link between the insecticide heptachlor epoxide, estradiol, and breast cancer, Cassidy [/bib_ref]. Another group evaluated breast cancer risk and exposure to environmental estrogens-more specifically, 16 organochlorine pesticides. The authors found that the presence of a higher concentration of the pesticides aldrin and lindane is associated with an increased risk of developing breast cancer, especially for leaner postmenopausal women. [bib_ref] Breast cancer risk and the combined effect of environmental estrogens, Ibarluzea Jm [/bib_ref] All these studies suggest an important role for pesticides and others environmental chemicals in the initiation and development of breast cancer; however, none of the studies thus far elude to mechanisms of action of these chemical, and how they affect the ERα and AhR signaling pathways (or their crosstalk).
Certain pesticides are capable of activating AhR and consequently inducing the expression of CYP1A1 and CYP1B1 genes. The enzymes encoded by these two genes are involved in 17β-estradiol (E2) metabolism: CYP1A1 and CYP1B1 convert E2 into 2-hydroxyestradiol (2-OHE2) and 4-hydroxyestradiol (4-OHE2), respectively [bib_ref] P450 enzymes of estrogen metabolism, Martucci [/bib_ref] [bib_ref] Cytochrome P450-mediated metabolism of estrogens and its regulation in human, Tsuchiya [/bib_ref]. Several studies have compared the tumorigenic potential of E2 and its metabolites, such as 2-OHE2 and 4-OHE2 [bib_ref] Molecular mechanisms of estrogen carcinogenesis, Yager [/bib_ref] [bib_ref] Induction of uterine adenocarcinoma in CD-1 mice by catechol estrogens, Newbold [/bib_ref] [bib_ref] Carcinogenicity of catechol estrogens in Syrian hamsters, Liehr [/bib_ref]. In fact, 2-OHE2 inhibits cellular growth of breast cancer cell lines [bib_ref] Estrogenic and antiestrogenic activities of 16α-and 2-hydroxy metabolites of 17β-estradiol in MCF-7..., Gupta [/bib_ref] , and induces apoptosis in immortalized mammary cells [bib_ref] 2-Hydroxyestradiol induces oxidative DNA damage and apoptosis in human mammary epithelial cells, Hurh [/bib_ref]. In contrast, 4-OHE2 induces kidney tumors in Syrian hamsters [bib_ref] Carcinogenicity of catechol estrogens in Syrian hamsters, Liehr [/bib_ref] , and enhances proliferation and mutagenesis by promoting the formation of depurinated adducts on DNA [bib_ref] Mutagenic activity of 4-hydroxyestradiol, but not 2-hydroxyestradiol, in BB rat2 embryonic cells,..., Zhao [/bib_ref]. 4-OHE2 is reported to be one of the most genotoxic estradiol metabolites. It has also been reported that breast cancer cells metabolize more 4-OHE2 than normal cells [bib_ref] Differential regulation of cytochrome P450 1A1 and 1B1 by a combination of..., Coumoul [/bib_ref]. Thus, the ratio between 2-OHE2/4-OHE2 metabolites, and consequently, the ratio between CYP1A1/CYP1B1 enzymes, appears to be important in the initiation of carcinogenesis in mammary tissues. In addition, if some pesticides induce more CYP1B1 expression than CYP1A1, this could also lead to imbalances in the ratio between CYP1A1 and CYP1B1 enzymes. Consequently, if we consider this scenario on a long-term exposure scale, more 4-OHE2 metabolites would be formed, hence leading to the accumulation of DNA adducts and potentially initiating cancer development [fig_ref] Figure 1: Proposed model for initiation of breast cancer by pollutants and pesticides [/fig_ref]. Estrogen metabolites have already been established as potential biomarkers for susceptibility to breast cancer [bib_ref] Relative imbalances in estrogen metabolism and conjugation in breast tissue of women..., Rogan [/bib_ref]. Thus, linking pesticide exposure to this imbalanced ratio in favor of the CYP1B1 enzyme would provide direct evidence that pesticides may play a role in the early steps of breast cancer development. Long-term exposure to pollutants and pesticides, which could induce the aryl hydrocarbon receptor (AhR) and the estrogen receptor α (ERα) signaling pathways, will create an imbalance between CYP1A1 and CYP1B1 enzymes. Thus, this will modify the 2-OHE2/4-OHE2 ratio and could contribute to mammary carcinogenesis.
It should also be taken into consideration that some of these pesticides probably contain impurities, and may even be contaminated with dioxins, which could only aggravate their adverse effects. The most notorious example is Agent Orange, an herbicide used during the Vietnam War by the U.S. military as a defoliating product. The now-banned herbicide was a mixture of 2,4-D (2,4-dichloro phenoxyacetic acid) and 2,4,5-T (2,4,5-trichlorophenoxyacetic acid), and it was later shown that 2,4,5-T was contaminated with small amounts of TCDD [bib_ref] Agent orange's bitter harvest, Stone [/bib_ref]. In 2004, the latest update of the Veterans and Agent Orange report of the National Academies' Institute of Medicine (IOM) claimed there was enough evidence of an association between exposure to this herbicide and the following illnesses: chronic lymphocytic leukemia, soft-tissue sarcoma, non-Hodgkin's lymphoma, Hodgkin's disease, and chloracne.
While the acute toxicity of pesticides has been well documented, there is still little known about the adverse effects of long-term chronic exposure on human health. According to the World Health Organization (WHO), long-term exposure to pesticides could increase the risk of developmental and reproductive disorders, immune-system disruption, endocrine disruption, impaired nervous-system function, and development of certain types of cancers, such as breast cancer. Attempts to establish correlations between the effects of pesticides on human health are particularly difficult because there are known sex, genetic, epigenetic, and environmental differences in the capacity to metabolize xenobiotics. Differences in susceptibility may also be affected by variations in the rate at which the xenobiotics are eliminated from the body. The adverse effects of pesticides may be related to their interactions with AhR, but they may also be partially or totally mediated by an AhR-independent signaling pathway. In both cases, specific windows of exposure to pesticides during a lifetime can have different outcomes thus complicating the interpretation of epidemiological studies. In addition, genetic differences in the properties of AhR are known to exist in human populations, and polymorphisms in cytochrome P450 enzymes have been associated with increased susceptibility to different cancers [bib_ref] Genetic polymorphism and toxicology-With emphasis on cytochrome p450, Johansson [/bib_ref].
Consequently, this demonstrates the challenges as well as the importance of better understanding the mechanisms that underline the crosstalk that exist between dioxin receptor and estrogen receptor signalling, and also rigorously test how different pollutants and pesticides affect this crosstalk.
## Molecular mechanisms of ahr and erα crosstalk
Interaction between the AhR and the ERα signaling pathways has been observed for several years. ERα belongs to the nuclear receptor family of transcription factors and is involved in the regulation of cellular proliferation in response to E2, for example during mammary gland development [bib_ref] Estrogen receptors: Structure, mechanisms and function, Hewitt [/bib_ref]. Numerous laboratories have focused their interests in studying the inhibition of the ERα signaling pathway by AhR. However, the role of ERα in the regulation of the expression of AhR target genes is less documented. Firstly, we will review the main conclusions and mechanisms proposed for the regulation of ERα by AhR. Secondly, we will discuss why the study of the differential regulation of AhR target genes (CYP1A1 and CYP1B1) by ERα is important, and why this can influence mammary carcinogenesis.
## Ahr-mediated repression of the erα signaling pathway
The earliest study to establish a link between pollutants and estrogen-induced cancer was carried out by . They made the striking observation that female Sprague-Dawley rats treated with TCDD for two years developed less mammary and uterine tumors than non treated rats [bib_ref] Results of a two-year chronic toxicity and oncogenicity study of 2,3,7,8-tetrachlorodibenzo-p-dioxin in..., Kociba [/bib_ref]. Ten years later, other studies showed that TCDD treatment also inhibits proliferation only in ERα positive breast cancer cell lines, and not in ERα negative cell lines [bib_ref] Inhibition of postconfluent focus production in cultures of MCF-7 human breast cancer..., Gierthy [/bib_ref] [bib_ref] Suppression of estrogen-regulated extracellular tissue plasminogen activator activity of MCF-7 cells by..., Gierthy [/bib_ref]. The chemopreventive and chemotherapeutic activities of TCDD in breast carcinogenesis triggered the interest of numerous laboratories to find or to develop AhR agonists that possess the antiestrogenic activity of TCDD, but without its acute toxicity. These AhR agonists are called Selective AhR Modulator (SAhRM). One example of such a SAhRM is the DIM, an acid-catalyzed metabolite of I3C, which is a compound found in cruciferous vegetables. Chen and colleagues showed that female rats treated with DIM had a decrease in E2-dependant 7,12-dimethylbenzanthracene (DMBA)-induced mammary tumors [bib_ref] Aryl hydrocarbon receptor-mediated antiestrogenic and antitumorigenic activity of diindolylmethane, Chen [/bib_ref]. However, DIM, additionally of AhR activation, also activates ERα by ligand-independent pathways mediated by the PKA and MAPK signaling [bib_ref] Potent ligand-independent estrogen receptor activation by 3,3'-diindolylmethane is mediated by cross talk..., Leong [/bib_ref] , which could have deleterious effects [bib_ref] Estrogen receptor alpha can selectively repress dioxin receptor-mediated gene expression by targeting..., Marques [/bib_ref].
The regulation of ERα by AhR acts at many levels, for which four different mechanisms have been proposed: firstly, studies showed that AhR and ERα interact with common transcription factors and coactivators in order to modulate transcription [bib_ref] Interactions of nuclear receptor coactivator/corepressor proteins with the aryl hydrocarbon receptor complex, Nguyen [/bib_ref] [bib_ref] Nuclear receptor coactivator SRC-1 interacts with the Q-rich subdomain of the AhR..., Kumar [/bib_ref] [bib_ref] Differential recruitment of coactivator RIP140 by Ah and estrogen receptors. Absence of..., Kumar [/bib_ref] [bib_ref] CBP/p300 functions as a possible transcriptional coactivator of Ah receptor nuclear translocator..., Kobayashi [/bib_ref]. Consequently, when the two pathways are activated simultaneously, AhR and ERα compete for the binding of these factors. Secondly, AhR represses some ER target genes by binding directly their promoter. The first example to document this mechanism was discovered at the cathepsin D promoter where the sequence of the pentanucleotide core of the XRE is found [bib_ref] Molecular mechanism of inhibition of estrogen-induced cathepsin D gene expression by 2,3,7,8-tetrachlorodibenzo-pdioxin..., Krishnan [/bib_ref] [bib_ref] Transcriptional activation of cathepsin D gene expression by 17β-estradiol: Mechanism of aryl..., Wang [/bib_ref]. The presence of the inhibitory XRE (iXRE) is necessary for the repression of cathepsin D by AhR. Functional iXREs were later identified in the promoter of c-fos, hsp27 and TFF1 genes [bib_ref] Transcriptional activation of c-fos protooncogene by 17β-estradiol: Mechanism of aryl hydrocarbon receptor-mediated..., Duan [/bib_ref] [bib_ref] Transcriptional activation of heat shock protein 27 gene expression by 17β-estradiol and..., Porter [/bib_ref] [bib_ref] Identification of a motif within the 5' regulatory region of pS2 which..., Gillesby [/bib_ref]. Thirdly, the activation of AhR increases the degradation of ERα by the ubiquitin-proteasome pathway [bib_ref] The aryl hydrocarbon receptor mediates degradation of estrogen receptor alpha through activation..., Wormke [/bib_ref] [bib_ref] Effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin and related compounds on the occupied nuclear estrogen receptor..., Harris [/bib_ref]. Ohtake and co-workers showed that AhR is associated with an E3-ubiquitine ligase complex, which is proposed to be involved in ERα degradation [bib_ref] Dioxin receptor is a ligand-dependent E3 ubiquitin ligase, Ohtake [/bib_ref]. Additionally, using an Estrogen response element (ERE) placed upstream of a luciferase reporter, they showed that the presence of MG132 (a proteasome inhibitor) abrogates the repression elicited by AhR in a ligand-specific manner. However, this result is challenged by other studies showing that MG132 affects the mRNA levels of the ERα target genes, even in the absence of AhR ligands [bib_ref] The 26S proteasome is required for estrogen receptor-α and coactivator turnover and..., Lonard [/bib_ref] [bib_ref] Cyclic, proteasome-mediated turnover of unliganded and liganded ERα on responsive promoters is..., Reid [/bib_ref]. The discrepancies between these results raise a major concern regarding data obtained with reporter constructs. Nevertheless, activated-AhR and E2-mediated degradation of ERα borrows two different pathways, which could explain why one is necessary for ERα target genes expression and the other one inhibits it [bib_ref] Dioxin receptor is a ligand-dependent E3 ubiquitin ligase, Ohtake [/bib_ref]. Fourthly, in breast cancer cells, TCDD induces the expression of CYP1A1 and CYP1B1, which encode enzymes that convert E2 into catecholestrogens [bib_ref] P450 enzymes of estrogen metabolism, Martucci [/bib_ref] [bib_ref] Cytochrome P450-mediated metabolism of estrogens and its regulation in human, Tsuchiya [/bib_ref]. The expression of these enzymes leads to an increase in E2 metabolism and a reduction of intracellular E2 concentration. Although, this mechanism may contribute in the repression of the ERα signaling pathway by AhR, it is not sufficient to explain all the cellular effects that are observed. Indeed, the repression of cathepsin D expression mediated by TCDD occurs very quickly after E2 treatment, and at this time point, CYP1A1
is not yet induced [bib_ref] Molecular mechanism of inhibition of estrogen-induced cathepsin D gene expression by 2,3,7,8-tetrachlorodibenzo-pdioxin..., Krishnan [/bib_ref]. Moreover, in rats treated with TCDD, the level of E2 circulating in the blood is not affected [bib_ref] Effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on serum concentrations and the uterotrophic action of exogenous..., Shiverick [/bib_ref]. In conclusion, the pathway involving E2 metabolizing enzymes is not necessary for the inhibitory crosstalk observed between AhR and ERα.
## Erα-mediated repression of ahr target genes
In mammary tissues, CYP1A1 and CYP1B1 are the two most induced AhR target genes after TCDD treatment. As previously mentioned, the enzymes encoded by these two genes are involved in E2 metabolism: CYP1A1 converts E2 in 2-OHE2, while CYP1B1 converts E2 in 4-OHE2, one of the most genotoxic estrogen metabolites. Interestingly, ERα specifically represses CYP1A1 gene expression, but not CYP1B1. We propose that action of ERα may constitute another mechanism by which it promotes carcinogenesis.
Our laboratory has focused efforts to unravel the mechanism(s) involved in the CYP1A1 gene repression by ERα in breast cancer cell lines. We discovered two new key players in the regulation of CYP1A1 expression in ERα positive cell lines: the histone variant H2A.Z and the DNA methyltransferase Dnmt3B. Depletion of H2A.Z in MCF7 cells triggers DNA methylation of AhR binding sites in the CYP1A1 promoter, and thus affects CYP1A1 induction in the presence of AhR agonists. In cell lines which do not express ERα, the absence of H2A.Z has no effect on the induction of CYP1A1 [bib_ref] Estrogen receptor alpha can selectively repress dioxin receptor-mediated gene expression by targeting..., Marques [/bib_ref].
Previous work from the Henikoff laboratory showed that H2A.Z antagonizes DNA methylation in Arabidopsis thaliana [bib_ref] Z and DNA methylation are mutually antagonistic chromatin marks, Zilberman [/bib_ref] and a similar conclusion has been drawn in a murine model [bib_ref] Changes in H2A.Z occupancy and DNA methylation during B-cell lymphomagenesis, Conerly [/bib_ref]. The same mechanism appears to be true at the CYP1A1 promoter in the presence of ERα. Moreover, we showed that ERα recruitment to the CYP1A1 promoter in presence of TCDD and E2 decreases AhR binding at its XREs, which leads to two times less induction of CYP1A1 mRNA level than in the presence of TCDD alone. We also discovered that the inhibition of DNA methylation, by either 5-azacytidine treatment or by depleting Dnmt3B, impairs specific inhibition of CYP1A1 by ERα [bib_ref] Estrogen receptor alpha can selectively repress dioxin receptor-mediated gene expression by targeting..., Marques [/bib_ref]. Taken together, these new findings suggest that DNA methylation plays a central function in the regulation of the CYP1A1 expression in the presence of ERα. A model is proposed for this mechanism in [fig_ref] Figure 2: Proposed model for the regulation of the CYP1A1 gene by ERα [/fig_ref]. In the absence of estradiol and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), CYP1A1 is not expressed and H2A.Z is present in its promoter. In the presence of both ligands, H2A.Z is removed and AhR/Arnt/ERα is recruited to the CYP1A1 promoter. ERα displaces AhR/Arnt by promoting DNA methylation on the XREs in the CYP1A1 promoter, thus resulting in less AhR activating surfaces available to stimulate CYP1A1 expression than in presence of TCDD alone.
# Conclusions
Industrialized countries generate more and more pesticides and pollutants. However, there is a general concern in fabricating useful chemicals that are not "as toxic" as older generation pesticides. Nonetheless, there is no way to tell how the pesticides used today in our agricultural practices will influence our health on a long-term basis. It is clear that even though a particular pollutant may not present a high health risk-probably based on toxicological tests performed with laboratory animals-it may still significantly perturb endocrine systems, and lead to the generation of genotoxic metabolites. We believe that genes involved in the regulation of the crosstalk that exists between the dioxin receptor and estrogen receptor signaling could become important molecular sensors, or biomarkers, to assess potential long-term effects of pesticides on certain forms of cancer.
[fig] Figure 1: Proposed model for initiation of breast cancer by pollutants and pesticides. [/fig]
[fig] Figure 2: Proposed model for the regulation of the CYP1A1 gene by ERα. [/fig]
|
A Collection of Pre-mRNA Splicing Mutants in Arabidopsis thaliana
To investigate factors influencing pre-mRNA splicing in plants, we conducted a forward genetic screen using an alternatively-spliced GFP reporter gene in Arabidopsis thaliana. This effort generated a collection of sixteen mutants impaired in various splicing-related proteins, many of which had not been recovered in any prior genetic screen or implicated in splicing in plants. The factors are predicted to act at different steps of the spliceosomal cycle, snRNP biogenesis pathway, transcription, and mRNA transport. We have described eleven of the mutants in recent publications. Here we present the final five mutants, which are defective, respectively, in RNA-BINDING PROTEIN 45D (rbp45d), DIGEORGE SYNDROME CRITICAL REGION 14 (dgcr14), CYCLIN-DEPENDENT KINASE G2 (cdkg2), INTERACTS WITH SPT6 (iws1) and CAP BINDING PROTEIN 80 (cbp80). We provide RNA-sequencing data and analyses of differential gene expression and alternative splicing patterns for the cbp80 mutant and for several previously published mutants, including smfa and new alleles of cwc16a, for which such information was not yet available. Sequencing of small RNAs from the cbp80 mutant highlighted the necessity of wild-type CBP80 for processing of microRNA (miRNA) precursors into mature miRNAs. Redundancy tests of paralogs encoding several of the splicing factors revealed their functional non-equivalence in the GFP reporter gene system. We discuss the cumulative findings and their implications for the regulation of pre-mRNA splicing efficiency and alternative splicing in plants. The mutant collection provides a unique resource for further studies on a coherent set of splicing factors and their roles in gene expression, alternative splicing and plant development.
## Arabidopsis
thaliana CBP80 miRNAs mutant screen pre-mRNA splicing
Splicing of pre-mRNAs by the excision of introns and ligation of flanking exons is a prerequisite for the expression of most eukaryotic genes. Splicing entails two transesterification reactions carried out by the spliceosome, a large and dynamic ribonucleoprotein (RNP) machine located in the nucleus. At least six structurally and functionally distinct spliceosomal complexes containing core spliceosomal proteins, transiently-associated factors and different combinations of five different small nuclear (sn) RNAs -U1, U2, U4, U5 and U6act sequentially to execute the two catalytic steps of the splicing process. The spliceosome is able to carry out constitutive splicing, in which the same splice sites are always used for a given intron, and alternative splicing, in which splice site usage for a given intron is variable. Alternative splicing increases transcriptome and proteome diversityand is important for development and stress adaptation in plants. Most information on spliceosome composition and the splicing mechanism has been derived from genetic, biochemical and structural studies in yeasts and metazoan cells . Structural and mechanistic insights into the splicing process in these organisms have relied heavily on the development of manipulable biochemical systems that perform the splicing reactions in vitro. Similar work has lagged in plants, largely owing to the lack of an efficient in vitro splicing system (although see recent progress in this area;. Many of the approximately 430 predicted splicing-related proteins encoded in the Arabidopsis thaliana (Arabidopsis) genome have been identified through sequence similarity searches with yeast, Drosophila and human genes. Genetic approaches have been useful for revealing the physiological roles for some of these splicing factors. For example, forward genetic screens designed to investigate requirements for distinct processes, such as hormone responsesand flowering time, have identified different splicing-related proteins. Reverse genetics has been used to study the consequences of specific splicing factor deficiencies at particular developmental stagesor under different environmental conditions (for example,. However, to our knowledge, an unbiased forward genetic screen dedicated to identifying factors that influence alternative splicing of a welldefined alternatively spliced gene has not previously been conducted in any plant system.
We developed a novel GFP reporter system in Arabidopsis to carry out such a forward genetic screen. In this system, an intron-containing GFP reporter gene present in a wild-type 'Target' (WT T) line exhibits variable levels of GFP expression depending on the splicing pattern of its pre-mRNA. Of three major GFP splice variants present in wild-type plants, only one, which arises from splicing a U2-type intron with comparatively weak, non-canonical AU-AC splice sites, produces a translatable GFP mRNA. The other two GFP transcriptsa spliced transcript resulting from excision of a U2-type intron with strong, canonical GU-AG splice sites, and an unspliced pre-mRNAcannot be translated because they comprise premature termination codons. Our working hypothesis is that mutations in genes encoding splicing factors will alter the relative proportions of the three splice variants, leading to an increase or decrease in the level of translatable GFP mRNA. Theoretically, these changes should result, respectively, in either a 'Hyper-GFP' (HGF) or 'GFP-weak' (GFW) phenotype compared to the WT T line, which displays an intermediate level of GFP fluorescence .
Using the GFP splicing reporter system, we conducted a classical forward genetic screen to identify mutants exhibiting altered splicing and GFP expression. By screening for changes in GFP fluorescence in seedlings derived from chemically mutagenized seed, we retrieved nine hgf mutants and seven gfw mutants. Eleven of the mutants, which indeed turned out to be defective in various splicing-related factors, have been published previously . Here we report the identity of the final five mutants together with previously unpublished information on global gene expression and alternative splicing profiles in one new mutant and in several mutants for which this information has not been previously published. Our study provides foundational knowledge for further in depth investigations of the splicing factors retrieved in the screen, and allows an integrated analysis of a set of sixteen splicing-related proteins unified by their involvement in processing the same splicing reporter.
# Materials and methods
# Plant materials
All wild-type and mutant plants used in this study are in the ecotype Col-0 background and were cultivated under long-day conditions (22-23°, 16 hr light, 8 hr dark).
The T-DNA insertion mutants used in this study were as follows: SAIL_569_G05 (rbp45a; AT5G54900), SAIL_505_E03 (rbp45b; AT1G11650), SALK_063484C (rbp45c; AT4G27000), SALK_ 009736C, SAIL_527_G04, SALK_132471, SALK_053475 and SALK_152624 (cwc16b; AT1G17130), SAIL_608_B05 (smfb; AT2G14285). All the seeds were obtained from the Nottingham Arabidopsis Stock Center (NASC).
## Forward genetic screen
The forward genetic screen based on an alternatively-spliced, introncontaining GFP reporter gene in Arabidopsis [T line; referred to herein as 'wild-type' (WT)]has been described in detail in prior publications . Briefly, approximately 40,000 Arabidopsis seeds of the WT T line homozygous for the alternatively-spliced GFP reporter gene were treated with ethyl methane sulfonate (EMS) and sown on soil (M1 generation). From approximately 30,000 M1 plants that grew to maturity and produced self-fertilized seeds, 52 batches of M2 seeds (the first generation when a recessive mutation can be homozygous) were harvested. Surface-sterilized M2 seeds were germinated on solid Murashige and Skoog (MS) medium in plastic Petri dishes and examined under a fluorescence stereo microscope at seven days post-germination for GFP fluorescence. M2 seedlings showing enhanced or reduced GFP fluorescence relative to the WT T line were placed into hgf and gfw categories, respectively, and selected for further analysis. Causal mutations were identified in the mutants by next generation mapping (NGM; James et al. 2013) using pooled DNA isolated from at least 50 BC1F2 progeny displaying the desired GFP phenotype. BC1F2 plants were produced by backcrossing the M2 plants with the WT T line followed by self-fertilization of the resulting BC1 progeny. Screening approximately 210,000 M2 seedlings (representing around seven M2 seedlings for each M1 plant) yielded nine hgf mutants and seven gfw mutants . Mutations were confirmed by identification of multiple alleles and/or complementation analyses. All mutations reported here are recessive.
## Complementation
For complementation of the rbp45d-1, dgcr14-1, cdkg2-3, iws1-2 and cbp80-1 mutants, the respective coding sequences (https://www.arabidopsis.org/) were modified to add SalI sites and XbaI at the 59 and 39 ends, respectively, and to replace internal SalI or XbaI sites by silent mutations. The modified cDNAs were synthesized by Genscript (www.genscript.com) and cloned into pUC57 as SalI-XbaI fragments.
The binary vector BV-MpPATot SalI(PAT, phosphinothricin acetyltransferase, which confers resistance to DL-phosphinothricin) was altered to contain the 35S promoter from cauliflower mosaic virus (35Spro)and 3C transcriptional terminator region (from the pea rbcS3C genebetween the PAT selection marker and the left T-DNA border region of the binary vector. Between the 35Spro and 3C terminator region, a chloramphenicol resistance marker was inserted between the SalI site after the 35Spro and the XbaI site positioned before the 3C terminator. By cutting with SalI and XbaI, the chloramphenicol resistance marker could be replaced by the modified CDS (under transcriptional control of the 35Spro) of genes to be complemented.
To permit immunoprecipitation of IWS1 in future experiments, the IWS1 cDNA (lacking a stop codon) was fused to three copies of a C-terminal FLAG tag. For this, an adaptor consisting of a SpeI fragment encoding three copies of a FLAG peptide sequence with a stop codon followed by XbaI NPTII XbaI SpeI fragment (NPTII, neomycin phosphotransferase conferring resistance to kanamycin) was synthesized by Genscript. This adaptor was inserted in the correct orientation into the XbaI site of the modified binary vector that contains an IWS1 cDNA lacking a stop codon, by selecting for kanamycin resistance in bacteria. The NPTII selection marker was deleted afterward with XbaI, thus reconstituting the 35Spro-IWS1-FLAG CDS.
Binary vectors containing the modified cDNAs were introduced into Agrobacterium tumefaciens strain ASEvia triparental mating and Arabidopsis plants were transformed using the floral dip procedure.
## Rna-sequencing and whole genome resequencing
Total RNA was isolated from approximately 80 mg of two week-old seedlings (BC1F3 generation) of each mutant line tested in this study (EMS-generated mutants: cbp80, smfa-1, cwc16a-1, cwc16a-2, cwc16a-3; SALK T-DNA insertion line: smfb) and wild type T plants using a Plant Total RNA Miniprep kit (GMbiolab, Taiwan) according to the manufacture's protocol for the Lysis Solution B, which contains SDS/anti-oxidant. Briefly, the seedlings were ground into a fine powder in liquid nitrogen, then lysed in the Lysis Solution B for 10 min at 60°, followed by column purification steps. The purified total RNA was extracted using nuclease-free water after the on-column DNase treatment. Construction of libraries and RNAseq were carried out (biological triplicates for each sample) as described previously.
Whole genome re-sequencing of the EMS-generated mutants cbp80, smfa-1, cwc16a-1, cwc16a-2, cwc1-a-3 was performed to identify any remaining EMS-induced second-site mutations that change splice sites. Alternative splicing events containing mutations were excluded from further analysis. To prepare the sample for whole genome re-sequencing, genomic DNA was isolated by DNeasy plant mini kit (Qiagen, Taiwan) followed by concentrating by The T-DNA construct introduced into Arabidopsis comprises a GFP reporter gene under the transcriptional control of a minimal promoter (TATA) and upstream viral (EPRV) enhancer. In the WT T line, however, the expected transcription initiation site (TSS) (light gray arrow) is not used. Rather transcription of GFP pre-mRNA initiates at an unanticipated upstream TSS (black bar and arrow). Alternative splicing yields three GFP splice variants: an unspliced transcript, a transcript resulting from splicing of a canonical GU-AG intron, and a transcript arising from splicing a U2-type intron with non-canonical AU-AC splice sites, which are considered weak compared to GU-AG splice sites. The unspliced and GU-AG transcripts contain numerous premature termination codons (black asterisks). Hence only the AU-AC transcript can be translated into GFP protein.
The coding sequence of GFP protein (green bars) uniquely contains a 27 amino acid extension (short green bars with black outline) compared to standard GFP. Arrowheads denote a tandem repeat cluster upstream of the cryptic promoter. The black AUG designates the major translation initiation codon. The 39 AC splice site is only three nucleotides downstream of the 39 AG splice site. (B) Schematic of forward screen and GFP fluorescence phenotypes of newly germinated seedlings. The wild-type Target (T) line has an intermediate level of GFP fluorescence visible primarily in the stem (hypocotyl) and shoot and root apices. Mutants generated by EMS treatment of the T line exhibit either reduced (Weak) or enhanced (Hyper) GFP fluorescence relative to the T line. Cotyledons (first leaves appearing from a germinating seedling) appear red owing to autofluorescence of chlorophyll at the excitation wavelength for GFP. Genomic DNA Clean & Concentrator (Zymo research, CA, USA) according to respective manufacturer's instructions. Briefly, about 100 mg of two-week-old seedlings were used as starting material and the purified DNA was eluted from the column by 200 mL of the elution buffer (provided by the kit). The DNA solution was then concentrated to 60 mL via column concentration steps. Construction of libraries and DNA-seq were carried out as described previously.
Analysis of RNA-seq data for differential gene expression and differential alternative splicing Differential expression analysis: To determine differential expression of the cbp80, coilin, cwc16a, and smfa and smfb mutants compared to their respective WT T controls, we estimated the transcript per million (TPM) expression with Salmon (version 0.13.1;for the Reference Transcript Dataset for Arabidopsis thaliana 2 (AtRTD2)-Quantification of Alternatively Spliced Isoforms (QUASI) (AtRTD2-QUASI) annotation. Transcript read counts were grouped per gene using tximportand differentially expressed genes were determined using edgeR with the exactTest (version 3.18.1;. Genes were considered differentially expressed for a false discovery rate (FDR) , 0.05.
Read alignment: Reads were mapped to the index based on the TAIR10 genome releaseand the AtRTD2 transcriptome with STAR (version 2.6.0c;using a 2-pass mapping. The following parameters were used:-outSAMpri-maryFlag AllBestScore,-outFilterMismatchNmax 2/0 (first/second pass),-outSjfilterCountTotalMin 10 5 5 5,-outFilterIntronMotifs RemoveNoncanonical,-alignIntronMin 60,-alignIntronMax 6000, n■ Mutants identified in a forward genetic screen based on an alternatively-spliced GFP reporter gene The mutants retrieved in a forward genetic screen based on an alternatively-spliced GFP reporter gene in Arabidopsisinclude a predicted core spliceosomal protein (SMFa); putative components of the U1 snRNP (PRP39a, RBM25, RBP45d) and U5 snRNP (PRP8); putative step I and step II factors transiently associated with the spliceosome (CWC16a and PRP18a, respectively); a predicted complex C protein (DGCR14); putative splicing regulatory proteins (RTF2, SMU1, PRP4ka, CDKG2); one structural protein presumed to be important for snRNP maturation (coilin), putative mRNA export factors (SAC3a, CBP80) and a predicted transcription elongation factor (IWS1). Developmental phenotypes are primarily observed in six (of seven) identified gfw mutations, two of which are embryo-lethal. a Further screening of the M2 population after publication of the first alleles of coilin, PRP39a and CWC16a has identified three new alleles of coilin (R9H; first intron, 39 splice site; second intron, 59 splice site), one new prp39a allele (R226 Ã ) and two new cwc16a alleles (W18 Ã ; fifth intron, 39 splice site). These unpublished alleles are counted in the number of alleles shown here. Abbreviation: SRA, Sequence Read Archive (NCBI); ABRC, Arabidopsis Biological Resource Center; T (or ST) refers to the WT T line harboring the alternatively-spliced GFP reporter gene. If the sequencing data from T line has a separate SRA number, it is noted in the table; n.d., not done; n.a., not available.
-outSAMtype BAM SortedByCoordinate. During the second pass, the splice junction files of the relevant control and test samples were passed to the mapping via the-sjdbFileChrStartEnd flag.
Alternative splicing analysis: Alternative splicing events were obtained and quantified using Whippet (version 0.11; Sterne-Weiler et al. 2018). Two separate splice graph indices were generated; one for the detection and quantification of exon skipping (ES), alternative acceptor (AA) and alternative donor (AD) events, and another for the retained introns (RI) and exitron (EI) events. Exitrons are alternatively spliced internal regions of protein-coding exons. Both indices were based on the AtRTD2 transcriptome annotation, supplemented with the relevant STAR RNA-seq alignments and were generated with the-bam-both-novel and-bam-min-reads 3 flags. The RI/EI index was further supplemented with 'pre-mRNA' coordinates of the genes and the exitron splice junctions detected using an in-house script. The 'pre-mRNA' coordinates range from the start to the end of the gene and allow us to quantify the retention levels of all annotated introns in a gene. The whippet delta step was run with default parameters, except for themin-samples 3 flag. The alternative acceptor and donor events were filtered, assuring that at least both (alternative) junctions were detected in the Whippet data. The RI events were required to be covered by at least one read for either all control and/or test samples. All events with a probability $ 0.9 and an absolute delta percentspliced-in (PSI) $ 0.1 were considered significant differential alternative splicing events.
SNP/indel calling: SNPs and indels were identified using the Genome Analysis Toolkit (GATK) pipeline (Van der Auwera et al. 2013). Picard (version 2.10.9, http://broadinstitute.github.io/picard) was used to generate the sequence dictionary for the TAIR10 genome release. Reads were aligned to the TAIR10 genome using BWA-MEM (0.7.16a-r1181; Li 2013), with the added -M flag. The resulting SAM file was converted to BAM format, sorted, and duplicates were marked using Picard tools. The GATK (version 3.8-0-ge9d806836) HaplotypeCaller was used to obtain the raw variants and the Select-Variants function was used to extract the SNPs and indels. SNPs were filtered using the following filter expression: "QD , 2.0 || FS . 60.0 || MQ , 40.0 || MQRankSum , 212.5 || ReadPosRankSum , 28.0." The filter expression for indels was as follows: "QD , 2.0 || FS . 200.0 || ReadPosRankSum , 220.0." SNPs and indels were intersected with the AtRTD2 annotated transcripts and the significantly regulated alternative splicing nodes from Whippet using in-house scripts.
RT-PCR to detect GFP splice variants: Total RNA was isolated as described above but without the on-column DNase treatment. Twenty-five microliters of the RNA solution were then treated with two units of RQ1 RNase-Free DNase (Promega, USA) in a total reaction volume of 50 ml according to the manufacturer's instructions. The DNase treated RNA was purified by NucleoSpin RNA Clean-up kit (Macherey-Nagel, Germany) and eluted with 60 ml of nuclease-free water. Following the manufacturer's protocol, cDNA was made by Transcriptor First Strand cDNA Synthesis Kit (Roche, USA) from 1 mg of the purified RNA and an oligo d(T) primer as a template and a primer, respectively. RT-PCR was carried out under the following conditions: [94°for 2 min followed by 28 cycles of 94°for 10 s, 58°for 20 s, and 72°for 90 sec, and finally 72°f or 7 min] or [94°for 2 min followed by 24 cycles of 94°for 10 s, 58°f or 20 s, and 72°for 30 s, and finally 72°for 7 min] for detecting GFP transcripts or Actin transcripts, respectively. Primers are shown in .
## Detecting gfp protein by western blotting
Approximately 100 mg of two-week-old seedlings grown on solid MS medium in plastic Petri dishes were frozen in liquid nitrogen, disrupted into a fine powder by TissueLyser II (Qiagen, USA) and resuspended in 100 ml of extraction buffer A (50 mM HEPES-KOH pH 7.9, 400 mM KCl, 2.5 mM MgCl2, 1 mM EDTA, 1 mM DTT, 0.1% Triton X-100) supplemented with EDTA-free protease inhibitor cocktail (Roche, USA). The suspension was vortexed three times for 15 sec and centrifuged for 10 min at maximum speed at 4°. The supernatants were mixed with equal volumes of extraction buffer A without KCl (sample A). The pellet was resuspended in 200 ml of extraction buffer B (50 mM HEPES-KOH pH 7.9, 200 mM KCl, 2.5 mM MgCl2, 1 mM EDTA, 1 mM DTT, 0.1% Triton X-100) supplemented with EDTA-free protease inhibitor cocktail (Roche, USA) and sonicated three times for eight seconds, 10% duty cycle and 20% power (Bandelin Sonoplus HD 2070 with MS 73 probe), followed by centrifugation for 10 min at maximum speed at 4°. The supernatants were mixed with sample A. Protein concentrations in the samples were measured using the Bradford assay. Five hundred nanograms of protein were separated by ExpressPlus PAGE Gel, 4-12% (Genescript, Taiwan), transferred to Amersham Hybond P 0.2 mm PVDF Membrane (GE Healthcare, USA), followed by Western blotting according to standard procedures. Rabbit anti-tubulin (AS10 680; Agrisera, Sweden), and mouse anti-GFP (CPA9022; Cohesion bioscience, Taiwan) antibodies were used at 1:1,000 dilutions. Secondary antibody, goat anti-rabbit IgG-conjugated with horseradish peroxidase (Agrisera, Sweden) and goat anti-mouse IgG-conjugated with horseradish peroxidase (Biorad, USA), were used, respectively, at a 1:10,000 dilution. The blots were developed using Amersham ECL Select Western Blotting Detection Reagent kit (GE Healthcare, USA).
## Pcr to identify gene knock-outs of t-dna insertion mutants
To isolate genomic DNA from the T-DNA insertion mutant lines, several two week-old seedlings were frozen in liquid nitrogen, disrupted into fine powder by TissueLyser II (Qiagen, USA) and resuspended in 250 ml of extraction buffer (100 mM Tris-HCl pH 8.0, 500 mM NaCl, 50 mM EDTA, 10mM b-mercaptoethanol). After adding 35 ml of 10% SDS, the samples were incubated for 10 min at 65°. After ammonium acetate precipitation and isopropanol precipitation followed by washing and drying steps, the DNA pellet was dissolved in an appropriate volume of TE buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA). Genotyping was carried out by PCR using specific primers listed in . PCR conditions were as follows [94°for 2 min followed by 35 cycles of 94°for 10 s, 58°for 20 s, and 72°for 1 min, and finally 72°for 7 min]. When genotyping by using CAPS or dCAPS methods, the PCR products were digested by a suitable restriction enzyme .
Small RNA analysis in the cbp80-1 mutant Total RNA was isolated from two-week old seedlings of the WT T line and cbp80-1 mutants using a PureLink Plant RNA Reagent (Thermo Fisher, USA) and MaxTract high-density gel tubes (Qiagen, USA) following the manufacturer's instructions. The quality and quantity of isolated RNA were checked with Agilent Bioanalyzer prior to usage. RNA concentrations were assessed by NanoDrop (ND-1000 spectrophotometer). Ten micrograms total RNA of each mutant line were used for library preparation and sRNA sequencing (Illumina HiSeq 2500 system) by an in house Genomic Technology Core Facility.
After the quality and adaptor trimming, the clean sRNA tags were processed according to a previously published procedure. In brief, the sRNA tags of 18-26 nt were mapped to the Arabidopsis genome (Release TAIR10)with Bowtie. Tags that exhibited more than 20 genomic hits or were mapped to chloroplast genome, mitochondria genome, rRNAs, tRNAs, snRNAs or snoRNAs were discarded. To carry out cross-library comparison, the read numbers are used with normalization in transcripts per fifty million (TP50M). The normalized read counts of tag sequences from individual sequencing libraries were calculated by dividing the raw value by the total abundance of adjusted total raw counts of each library, and then multiplied by 50 million.
Northern blotting of sRNA was performed as described previouslywith a few modifications. Briefly, ten micrograms of total RNA were separated on 10% denaturing polyacrylamide TBE-Urea gels (Thermo Fischer, USA) and transferred to Hybond-N+ membranes (GE Healthcare, USA) using Electro Blot Mini System (Major Science, Taiwan). The membrane was then UV cross-linked with 120-mJ energy and baked for 1 h at 80°. DNA oligonucleotides complementary to miRNA (miRBase version 21)were used as probes to determine the expression of miRNA (sequences of probes in legend of . The probes were g-32 P end-labeled using T4 polynucleotide kinase (New England Biolabs, USA). Probe hybridizations were performed using Ultrahyb-Oligo buffer (Thermo Fischer, USA) at 37°overnight. After washing with buffer containing 2· SSC and 0.1% SDS, the membranes were exposed on Phosphor-Imager screens, and scanned using the Typhoon Scanner (GE Healthcare, USA). These membranes were also exposed to X-ray film for 1-7 days.contains data on the rbp45d mutants; : contains data on the dgcr14 mutants contains data on the cdkg2 mutant; contains data on the iws1 mutant; contains data on the cbp80 mutant; contains a Northern blot analysis of selected miRNAs in the cbp80-1 mutant; shows data on a redundancy test of SMFA and SMFB; contains data on the cwc16a mutants; CWC16; Table S1 contains primer sequences;contains comparative phenotypic data on the cbp80-1 mutant, cbp80 complemented plants and wild-type; contains an analysis of differential alternative splicing events in the cbp80 mutant; contains an analysis of differentially expressed genes and miRNAs in the cbp80 mutant; contains an analysis of differential alternative splicing events in the coilin mutants; contains an analysis of differentially expressed genes in the coilin mutants; contains an analysis of differential alternative splicing events in the cwc16a mutants; contains an analysis of differentially expressed genes in the cwc16a mutants;contains an analysis of differentially expressed genes in the smfa and smfb single mutants; contains an analysis of differential alternative splicing events in the smfa and smfb single mutants; contains an analysis of differential alternative splicing events in the smfa smfb double mutant;contains an analysis of differentially expressed genes in the smfa smfb double mutant.
## Data availability
Seeds of all mutant and wild-type Arabidopsis thaliana lines listed in are available at the Arabidopsis Biological Resource center (ABRC, Ohio, USA) and all DNA and RNA sequence data for selected mutants and the wild-type T line are available at NCBI under the respective accession numbers listed as follows. hgf1/ coilin/At1g13030/ ABRC stock number: CS69632, CS69639; NCBI accession numbers: R40 Ã /hgf1-1 and P439L/hgf1-8: SRP071829, T line: SAMN12817582, P439L/hgf1-8:SAMN12817583, R40 Ã /hgf1-1: SAMN12817584, T line, P439L/hgf1-8, and R40 Ã /hgf1-1: SRP089352 and SRP089656. hgf2/CWC16a/At125682/ ABRC stock number: CS69846, CS72366, CS72367; NCBI accession numbers: this study, cwc16a-1: SRP093582, T line: SAMN12817589, cwc16a-1: SAMN12817590, cwc16a-2: SAMN12817591, cwc16a-3: SAMN12817592. hgf3/SMU1/At1g73720/ ABRC stock number: N623852; NCBI accession numbers: smu1-1: SRP093582. hgf4/ SMFA/At4g30220/ ABRC stock number: CS69848; NCBI accession numbers:
# Results
The sixteen factors identified in the forward genetic screen are listed in . All correspond to putative splicing-related proteins predicted to act at a number of steps of the spliceosomal cycle and snRNP biogenesis pathway , respectively). More than one allele was retrieved for ten of the mutants, suggesting that the screen is close to saturation. Eleven of the mutants have been reported previously. The published hgf mutants include coilin/hgf1; cwc16a/hgf2, smu1/hgf3, smfa/hgf4; and prp39a/hgf5. The gfw category includes rtf2/gfw1 and prp8a/gfw2; rbm25/gfw3; prp18a/gfw4; and prp4ka/ gfw5 and sac3a/gw6. Here we present the final five mutants: four in the hgf category (rbp45d, dgcr14, cdkg2, and iws1) and one in the gfw group (cbp80).
## New hgf mutants
rbp45d/hgf6 -At5g19350: NGM analysis of two mutants placed in the hfg6 group revealed mutations in the gene encoding RBP45D (RNA-Binding Protein 45D). Arabidopsis RBP45D, which is 425 amino acids in length, is a nuclear RNA binding protein that contains three RRM (RNA Recognition Motif) domains and preferentially binds to nuclear polyA + RNA in vitro. The two rbp45d mutations that we recovered result, respectively, in a premature termination codon (W306 Ã ; rbp45d-1) and a frame shift caused by a mutation in the splice site acceptor of the 6 th (out of seven) intron (rbp45d-2).. A wild-type copy of the RBP45D gene complemented the hyper-GFP phenotype when introduced into an rbp45d mutant. The two rbp45d mutants did not show any obvious growth or developmental abnormalities.
The RBP45D orthologs in budding yeast and humans -Nam8p and TIA-1, respectivelyare U1 snRNP components that stimulate splicing at weak splice sites. Although RBP45D has been shown previously to lack this capability, an Arabidopsis protein similar to RPB45D, termed UBP1, appears to enhance splicing at otherwise inefficiently spliced introns. In the alternatively-spliced GFP reporter system, somewhat increased levels of both the translatable GFP AU-AC transcript, which may contribute to the hyper-GFP phenotype of the rpb45d mutant, and the non-translatable GU-AG transcript were observed. These results might indicate that wild-type RBP45D can, in some cases, suppress splicing at both weak and strong splice sites. Another possibility is that mutations in RBP45D contribute to a hyper-GFP phenotype through a mechanism other than splicing regulation. Further work is needed to investigate these possibilities.
RBP45D has three paralogs in Arabidopsis: RBP45A (At5g54900), RBP45B (At1g11650) and RBP45C (At4g27000). All four RBP45 genes are ubiquitously expressed, although RBP45D is expressed at a lower level than its three paralogs (http://bar.utoronto.ca/). We tested whether mutations in RBP45A, RBP45B, and RBP45C would affect expression of the GFP reporter gene by crossing the respective T-DNA insertion mutants with the WT T line. The F2 seedlings descending from these crosses were genotyped for homozygosity of the desired T-DNA insertion mutation. Unlike the rbp45d mutants, the rbp45a, rbp45b or rbp45c mutants, which represent T-DNA knockout insertions, did not display a hyper-GFP phenotype or increased accumulation of GFP protein relative to the WT T lineand S1E, respectively). The RBP45 paralogs are thus functionally nonequivalent in splicing in the GFP splicing reporter system. dgcr14/hgf7 -At3g07790: NGM analysis of two mutants placed in the hfg7 group revealed mutations in the gene encoding DGCR14-like [DiGeorge Syndrome Critical Region, also termed ESS2 (Expression Studies2) in humansand EsS-2 (ES2-Similar) in Caenorhabditis elegans]. DiGeorge Syndrome is a human pleiotropic developmental disorder that is caused by a chromosome 22 deletion, the shortest of which contains the DGCR14 gene. DGCR14-related proteins, which belong to the ESS2 superfamily of proteins, are evolutionarily conserved from fission yeast to humans. ESS2 proteins are typically around 500 amino acids in length and contain two predicted coiled-coil domains but no other recognizable functional domains.
There are not yet any published reports on DGCR14-like proteins in plants. In Arabidopsis, DGCR14-like is a single copy, intronless gene that encodes a protein 509 amino acids in length. We recovered two alleles in the screen: dgcr14-1 (Q80 Ã ) and dgcr14-2 (W365 Ã ) . The corresponding mutants show a hyper-GFP phenotype and increased accumulation of GFP protein, which can be complemented with a wild-type copy of the DGCR14-like gene . No obvious growth, morphological or reproductive defects were observed in the dgcr14 mutants. DGCR-like is ubiquitously expressed in Arabidopsis (http://bar.utoronto.ca/) and predicted to encode a nuclear-localized protein (http://suba.live/).
From biochemical studies in human cells, DGCR14 appears to be most abundant in the spliceosomal C complex . This affiliation remains to be confirmed in plants. The mechanism by which DGCR14 acts in splicing is not yet known. In C. elegans, ESS-2 has been found to foster accurate mRNA splicing when a splice site contains non-canonical sequences. The increased level of the translatable AU-AC GFP transcript in dgcr14 mutants suggests that the wild-type DGCR14 protein can suppress splicing at weak splice sites , but additional experiments are required to confirm this idea.
cdkg2/hgf8 -At1g67580: NGM analysis of the hgf8 mutant identified a mutation in the gene encoding CDKG2 (Cyclin-Dependent Kinase G2). CDKs constitute an evolutionarily conserved group of serine/ threonine kinases that have diverse roles in eukaryotes including cell cycle regulation, transcriptional modulation, pre-mRNA splicing, and translation. In humans CDKG11, which together with CDKG10 is most closely related to plant CDKGs, is particularly noted for its role in splicing. Arabidopsis CDKG2 has not yet been implicated in mRNA processing; however, its close homolog, CDKG1 (At5g63370), was reported to associate with the spliceosome and regulate splicing of a gene involved in pollen wall formation. CDKG1 was also found to modulate temperaturesensitive alternative splicing of a factor involved in regulation of flowering time in a pathway that involves CDKG2 in the thermosensing mechanism.
CDKG2 is a ubiquitously expressed, nuclear protein that is 752 amino acids in length (http://bar.utoronto.ca/; http://suba.live/). Two T-DNA insertion alleles, cdkg2-1 and cdkg2-2 (SALK_012428 and SALK_090262) have been published previously. The allele we isolated in the screen, cdkg2-3, results in a D530N substitution that alters a highly conserved amino acid in the highly conserved kinase domain . The position of the cdkg2-3 mutation suggests the kinase activity of CDKG2 is important for its function in splicing in the GFP reporter gene system but this remains to be tested in the future. The level of the translatable AU-AC GFP transcript increases substantially in the cdkg2-3 mutant, which presumably accounts for the hyper-GFP phenotype of the mutant . By contrast only trace amounts of the non-translatable GU-AG and unspliced transcripts can be detected by RT-PCR . The cdkg2-3 mutant has a normal appearance but it is somewhat early flowering (approximately four days before WT plants under our long-day growth conditions), as has been shown previously with the cdkg2-1 and cdkg-2 T-DNA insertion alleles mentioned above.
iws1/hgf9 -At1g32130: The hgf9-1 mutation was identified by NGM as an allele of the evolutionarily conserved transcription factor IWS1 (Interacts With Spt6). In yeast, Iws1 has been shown to act in a complex with RNA polymerase II and Spt6 to facilitate pre-mRNA splicing, efficient mRNA export, and transcription elongation accompanied by histone H3K36 methylation, which is a mark of transcriptionally active chromatin.
The IWS1 protein in Arabidopsis, which is 502 amino acids in length, has a negligibly-expressed paralog, IWS2 (At4g19000). Two alleles of iws1, iws1-2 and iws1-3, were recovered in the screen. The iws1-2 mutation alters the acceptor site of the 9 th intron, which is the penultimate intron, and the mutation in iws1-3 results in a P446L amino acid substitution . The hyper-GFP phenotype of iws1-2 can be complemented by a wild-type copy of the IWS1 gene . The iws1-2 mutant plants are viable and do not display any obvious defects in growth, development or reproduction under normal growth conditions. Mutations in IWS1 were also isolated in independent forward screens in Arabidopsis for factors required for brassinosteroid-induced gene expression (seb1; suppressor of bes1-d)and for High Nitrogen Insensitive (HNI) plants. IWS1 thus appears to participate in multiple physiological processes in plants, which probably reflects its predicted participation in transcription elongation, pre-mRNA splicing, epigenetic modulation, and mRNA export.
It is unclear how the iws1 mutations we identified confer a hyper-GFP phenotype. Epigenetic modifications, including various histone marks, are known to affect splicing efficiency. Conceivably, IWS1-dependent histone modifications in the GFP transcribed region could be altered in the iws1 mutants, thus potentially affecting splicing of GFP pre-mRNA. However, the splicing pattern of GFP pre-mRNA changes only slightly in the iws1-2 and iws1-3 mutants . An alternative role for IWS1 in modulating transport of GFP mRNA from the nucleus to the cytoplasm is conceivable but remains to be further examined in the context of the hyper-GFP phenotype of the iws1 mutants.
## New gfw mutant
cbp80/gfw7 -At2g13540: NGM analysis of the gfw7 mutant identified a mutation in the gene encoding CBP80 (Cap Binding Protein 80). CBP80, together with CBP20, forms the heterodimeric cap binding complex (CBC), which binds the 7-methylguanosine cap at the 59-end of eukaryotic mRNAs. The CBC participates in multiple processes in the cell, including transcription, splicing, transcript export, and translation and . The CBC was initially isolated biochemically from HeLa cells, where it was shown to be important for splicing. In Arabidopsis, the ortholog of CBP80 was first identified genetically in a forward screen Spliceosomal cycle and factors identified in the forward genetic screen. Splicing is catalyzed by the spliceosome, a large and dynamic ribonucleoprotein (RNP) machine located in the nucleus. Spliceosomes comprise five small nuclear (sn) RNPs, each containing a heptameric ring of Sm or Like-Sm proteins and a different snRNA (U1, U2, U4, U5 or U6), as well as numerous other non-snRNP proteins. During the splicesomal reaction cycle, the five snRNPs act sequentially on the pre-mRNA with a changing assemblage of non-snRNP proteins to form a series of complexes that catalyze two consecutive trans-esterification reactions. In complex E, U1 and U2 snRNPs first recognize the 59 and 39 splice sites branch points of introns and interact to form pre-spliceosomal complex A. The subsequent addition of preformed U4/U5/U6 tri-snRNP creates pre-catalytic complex B. Ensuing reorganization steps induce release of U1 and U4 snRNPs and conversion of complex B to complex B Ã , which catalyzes the first reaction yielding the free 59 exon and lariat 39-exon intermediates. The newly formed C complex catalyzes the second reaction to achieve intron lariat excision and exon ligation. Lastly, dismantling of the spliceosome frees individual components to assemble anew at the next intron. The positions of factors identified in our screen, as predicted by their orthologs in yeast or metazoans, are indicated by colored rectangles (green, Hyper-GFP; dull green, GFP-weak). Adapted by permission from Springer Nature, Nat. Rev. Mol. Cell Biol. 15: 108-121, A day in the life of the spliceosome, A. G.for abscisic acid (ABA) hypersensitive (abh1) mutants.
The Arabidopsis CBP80 protein is 848 amino acids in length. The mutation identified in this screen, cbp80-1 (= abh1-9), creates a premature termination codon (W630 Ã ) in the middle of a conserved MIF4G (Middle domain of eIF4G), type II domain . It is not clear which step(s) in GFP pre-mRNA processing and transport is affected in the cbp80-1 mutant. A role in splicing is suggested by an increase in the untranslatable, unspliced GFP transcript and decrease in the translatable GFP transcript resulting from splicing at the AU-AC splice sites in the cbp80-1 mutant . However, one or more other CBP80-dependent steps important for GFP snRNP maturation pathway and factors identified in the forward genetic screen snRNAs are transcribed by RNA polymerase II in the nucleus and transported to the Cajal Body (CB), where they undergo quality control steps before export to the cytoplasm. In the cytoplasm, the monomethyl guanosine 59 cap is converted to trimethyl guanosine by the enzyme TGS1. The snRNAs are encircled by a heteromeric ring complex comprising seven Sm core proteins, which protect the snRNA and together with the trimethyl guanosine cap, act as nuclear import signals. Back in the nucleus, the snRNP transits again through the CB where it undergoes further maturation steps before being released into the nucleoplasm to assemble into spliceosomes. Predicted positions of factors identified in our screen are shown by colored rectangles (bright green, Hyper-GFP; dull green, GFP-weak). Adapted by permission from Springer Nature, Nat. Rev. Mol. Cell Biol. 15: 108-121, A day in the life of the spliceosome, A. G.expression could also be affected . Further work in the future is needed to investigate this question.
In addition to its GFP-weak phenotype , the cbp80-1 mutant has a visible developmental phenotype featuring serrated rosette leaves, decreased rosette diameter, short stature and reduced seed set ,. These phenotypes, some of which have been noted previously in other cbp80 mutants, can be at least partially complemented by introducing a wild-type copy of the CBP80 gene into the cbp80-1 mutant, as can the GFP-weak phenotype and S5D;.
To determine the effects of the cbp80-1 mutation on global gene expression and alternative splicing patterns, we carried out triplicate RNA-seq and analyzed the sequencing data for differentially expressed genes (DEGs) and differential alternative splicing (DAS) events in the cbp80-1 mutant. In agreement with earlier results obtained from another cbp80 mutant using an RT-PCR alternative splicing panel, we found that a number of transcripts are differentially spliced in the cbp80-1 mutant. A notable feature of the DEG list in the cbp80 mutant is the increased accumulation of 53 microRNA (miRNA) precursors (pri-miRNAs) , sheet miRNA). The high representation of pri-miRNA-encoding genes in the list of up-regulated DEGs was reflected in the GO analysis, which identified highly significant increases in the expression of genes involved in gene silencing by miRNAs and posttranscriptional regulation , GO_UP_P sheet). Subsequent sequencing of small RNAs from the cbp80-1 mutant demonstrated that the heightened accumulation of pri-miRNA transcripts was paralleled by a decrease in many of the corresponding mature miRNAs . These results expand on prior findings from tiling array experiments, which demonstrated that CBP80 is required for pri-miRNA processing.
## Redundancy tests
We previously carried out tests for functional redundancy of paralogs of PRP39A, PRP18A and PRP4KA, which were all identified in the GFP splicing reporter screen . Similar to results obtained for RBP45D paralogs (see above), none of the paralogs tested previously (PRP39B, PRP18B and PRP4KB) were functionally equivalent to the A forms of the respective genes . In the present study, we tested SMFB for functional redundancy with its paralog, SMFA, which was identified in the screen as the hgf4-1 mutant (P16 Ã ) (Table 1; and S7B). A knockout T-DNA insertion allele of smfb was introduced into the wildtype T line by crossing and F2 seedlings descending from the resulting F1 plants were screened for homozygous smfb progeny in a homozygous TT background. Unlike smfa homozygous seedlings, homozygous smfb seedlings displayed neither a hyper-GFP phenotype nor increased amounts of GFP protein . Moreover, in contrast to smfa, which accumulated elevated levels of the translatable AU-AC GFP transcript relative to WT plants, the GFP pre-mRNA splicing pattern in the smfb mutant was virtually unchanged from the WT T line . We thus conclude that SMFA and SMFB are functionally nonequivalent in the GFP reporter gene system.
A mutation in the gene encoding putative step 2 factor CWC16a was initially identified in the GFP splicing reporter screen as hgf2/cwc16a-1 (F50 Ã ) and two new alleles, cwc16a-2 (W18 Ã ) and cwc16a-3 (frame shift from amino acid 194), are reported here . Attempts to test mutations in CWC16B, the paralog of CWC16A, were unsuccessful because of the five T-DNA insertion lines of cwc16b available from seed stock centers, three did not contain the T-DNA at the expected location and two did not appear to be cwc16b knockout alleles . However, the fact that we recovered three alleles of cwc16a and none of cwc16b in our screen suggests a special role for CWC16A in splicing GFP pre-mRNA splicing. Indeed, the levels of the translatable AU-AC GFP transcript and the amount of GFP protein increase substantially in the cwc16a mutants and S8D, respectively), which is fully consistent with their strong hyper-GFP phenotypes . The lack of a convincing cwc16b knockout mutation prevented testing the splicing pattern of GFP pre-mRNA in a cwc16b mutant and the viability of a double cwc16a cwc16b mutant.
Additional RNA-seq analyses from previously published mutants We previously published RNA-seq data and analyses of DEGs and DAS events in two coilin mutants [(hgf1-1 (R40 Ã ), hgf1-8 (P439L)] n■and one allele of cwc16a (cwc16a-1). However, these earlier analyses lacked either an optimal number of biological replicates (coilin alleles;or were performed on only a single allele (cwc16a-1;. Here we expand on these prior results by reporting new RNA-seq data (in triplicate) and DEG and DAS analyses from the two coilin alleles mentioned above and from two newly identified alleles of cwc16a: cwc16a-2 and cwc16a-3 . The more comprehensive results obtained by using multiple alleles of each mutant can be compared and lists of DEGs and DAS events shared by all alleles compiled. These comparisons hone the findings and provide a more accurate assessment of genes that robustly change in expression and alternative splicing profiles in a given mutant. The total number of shared DAS and DEG events among multiple alleles is shown in, with details, respectively, in Tables S5 and S6 (coilin/hgf1 alleles) and Tables S7 and S8 (cwc16a alleles). A notable feature of the coilin DEG data are the highly significant representation of genes involved in metabolism of secondary compounds and responses to various stresses and environmental stimuli , GO_share_UP_P). A previous albeit more limited transcriptome analysis of coilin mutants similarly suggested a prominent role for coilin in stress responses. We also present here new triplicated RNA-seq data and DEG/ DAS analyses for smfa-1, smf-1b and smfa-1 smf-1b double mutants. This information was not reported in a previous publication on smfa-1 mutants. An analysis of the RNA-seq data obtained with smfa-1 or smfb-1 single mutants revealed about twice as many DEGs (FDR , 0.05) in smfa-1 compared to smfb-1 (5058 and 2392, respectively, with 827 shared between the two paralogs). More than 100 DEGs in smfa-1, compared to only seven in smfb-1, correspond to known or predicted splicing factors in Arabidopsis (numbering around 430;share and b_Koncz share). These results suggest that wild-type SMFA-1 is involved more frequently than SMFB-1 in modulating the expression of splicing-related factors. When considering DAS events, the difference between the two paralogs is even more pronounced (approximately sixfold). We identified 2184 cases of DAS for smfa-1 vs. 353 for smf-1b, with 98 instances of DAS shared between the two paralogs. The limited overlaps in DEGs and DAS events in the smfa-1 and smf-1b mutants suggest that SMFA-1 is the predominant paralog acting in both gene expression and in splicing in Arabidopsis. These findings extend to the genomewide level the functional non-equivalence that was observed with the two smf paralogs in the GFP splicing reporter system. Double smfa-1 smfb-1 mutants are viable, and information about DAS and DEG events in the double mutant, which are essentially a sum of the individual smfa-1 and smfb-1 results, is shown in Tables S11 and S12, respectively.
# Discussion
In a forward genetic screen designed to identify factors involved in alternative splicing of a GFP reporter gene in Arabidopsis, we recovered sixteen splicing-related proteins that are predicted to have a variety of roles in the spliceosomal cycle and snRNP biogenesis pathway . Based on the functions of their orthologs in other organisms, some of the factors we identified are likely to be components of a particular snRNP (U1: PRP39A, RBM25, RBP45D; U5: PRP8A). Others are predicted to be associated with a specific spliceosomal complex (complex C: DGCR14), or to be required at a specific catalytic step of splicing (step 1: CWC16A; step 2: PRP18A). Several factors are potentially splicing regulators involved in catalyzing or targeting various post-translational modifications, including protein phosphorylation (PRP4KA, CDKG2) and ubiquitination (RTF2, SMU1). We also identified factors presumed to be important for snRNP biogenesis (SMFA, coilin); mRNA transport (SAC3A); and transcription elongation/histone methylation (IWS1). One factor, CBP80, is potentially involved in multiple steps including splicing in the nucleus, mRNA export, and snRNA maturation in the cytoplasm. The data we gathered have verified a role for a number of previously uncharacterized proteins in pre-mRNA splicing in plants and revealed novel morphological and developmental phenotypes conferred by specific mutations. Our work has also generated new information on genome-wide gene expression and alternative splicing profiles of endogenous genes in the respective mutants. The overall findings provide foundational knowledge that can underpin more in-depth investigations of the splicing-related proteins in the future.
In addition to its foundational aspects, our study is strengthened and expanded in scope by merging and comparing data from the complete collection of mutants. A combined analysis permits a broader understanding of alternative splicing that cannot be gleaned from the examination of single mutants alone. For example, an overview of the current data set already illustrates the extraordinary complexity of splicing regulation. Even though different mutations often have the same effect on GFP pre-mRNA splicing and GFP expression, there is little overlap in the population of endogenous genes exhibiting splicing defects in the respective mutants. Substantial overlap would be expected if the mutations were revealing distinct, common features of introns affected in both mutants. Therefore, it has not been possible to discern fixed rules that govern splicing from the extant data. The findings rather indicate that each intron represents a unique context for splicing to occur, such that the effects of specific mutations on the splicing outcome are largely unpredictable. The application of various post-translational modifications to splicing factors, which is suggested by our identification of putative protein kinases and ubiquitination targeting factors in the screen, adds a further significant dimension to the regulation of alternative splicing that is only beginning to be understood (de la Fuente van.
Another interesting point arising from the cumulative findings is that obvious developmental phenotypes are observed primarily with GFP-weak mutants . Although the biological significance of these findings is still uncertain, they may reflect different developmental roles for proteins required for the splicing reactions to take place at all vs. proteins that influence splice site selection but do not affect the occurrence of splicing per se. The latter category may be more important in plants for modulating splicing patterns in response to stress conditions or other transitory signals that induce an adaptive response.
Unlike their counterparts in yeasts and metazoans, many genes encoding splicing factors in Arabidopsis and other higher plants are duplicated but the extent of functional redundancy or functional divergence has not been clear in most cases. In our mutant collection, we found a number of cases in which paralogs have a non-redundant function in GFP pre-mRNA splicing . Paralogs of genetically-identified factors that we tested directly and found to be non-redundant in the GFP reporter gene system include PRP39B, PRP4KB, PRP18B, SMFB, and RBP45A, B and C. Although a direct examination of CWC16B was not possible owing to the lack of a suitable knockout T-DNA insertion allele, we presume that CWC16A is the main form involved in splicing GFP pre-mRNA because our screen identified three alleles of cwc16a and none of cwc16b. Similarly, SAC3A and PRP8A were the only paralogs of the respective genes retrieved in the screen, suggesting a specific role for the A forms in the GFP splicing reporter system. . Although null mutations in these factors may be lethal, weak mutations that affect GFP splicing and expression could conceivably have been identified in the screen. For example, weak alleles of rtf2 and prp8 were recovered in the screen even though null mutations in these factors are embryo-lethal. Analysis of the RNA-seq data for both A and B forms of SMF revealed only modest overlaps in DEGs and DAS events between the paralogs, extending the functional non-equivalence of SMFA and SMFB that was observed with the GFP reporter gene to the genomewide level. Extensive functional divergence of paralogous plant genes encoding splicing factors increases the complexity and capabilities of the splicing machinery and is likely to be another feature of splicing in plants that contributes to their developmental plasticity and ability to adjust to a constantly changing environment.
Another noteworthy aspect emerging from the combined analysis is that mutations were retrieved in splicing factors predicted to act throughout the splicesomal cycle and snRNP maturation pathways and not just at a single crucial step of GFP pre-mRNA splicing. The only possible bias concerns predicted constituents of the U1 snRNP, which is involved in recognizing the 59 splice site early in the spliceosomal cycle. Of these, we identified three putative U1 snRNA components in the screen: RBP45D, PRP39A and RBM25 . The otherwise widespread distribution of the identified mutations as well as the participation of only single members of paralogous gene pairs at different steps of the splicing process hint that our system is illuminating a specialized pathway comprising a set of dedicated components.
The features of the alternatively-spliced GFP reporter gene that permit the genetic identification of a coherent set of splicing factors that cooperate in splicing of GFP pre-mRNA are not known. Potentially, however, they could be a subject for future investigation using CRISPR-Cas9-mediated genome editing to alter nucleotides within the spliced regions of the GFP gene. To our knowledge, many of the splicing-related proteins we identified have not been picked up in any other forward screen focusing on any type of process in plants, which again points toward one or more distinctive but as yet unidentified features of our GFP splicing reporter that render it particularly sensitive to mutations in the genetically identified factors. Based on the strong preference of their cognate mutants for the weaker AU-AC splice sites in GFP pre-mRNA, some HGF factors, such as CWC16A, SMU1, DGCR14 and CDKG2, might be important for discriminating between strong and sub-optimal/weak splice sites in the GFP pre-mRNA. Whether this observation extends to endogenous genes remains to be determined by scrutinizing in more detail the genome-wide RNA-seq data to assess both canonical and noncanonical splice site usage.
In summary, we used a unique alternative splicing system with both strong (GU-AG) and weak (AU-AC) splice sites, and an easy readout (GFP expression) to identify a coherent set of splicing factors that act at different stages of GFP pre-mRNA splicing and expression in Arabidopsis. The GFP pre-mRNA system is versatile and, similarly to other plant splicing reporters, can be used for other purposes, for example, to test mutations in other splicing factors or under different growth/environmental conditions. The resources generated in this study, including seeds of all mutants and high-throughput sequencing data, have been deposited in public repositories and hence are available to the international plant science community. We anticipate that additional global insights into the regulation of alternative splicing in plants will emerge once an entire data set containing detailed phenotypic and RNA-seq data from all mutants is available. |
A Case of Myopericarditis After the Second Dose of mRNA COVID-19 Vaccine in a Patient With a History of Myopericarditis
# Introduction
The clinical course of myocarditis varies from mild to fulminant. The onset of myocarditis often follows a viral illness. In addition, certain nonviral infections and autoimmune syndromes remain important causes of myocarditis. [bib_ref] The quest for new approaches in myocarditis and inflammatory cardiomyopathy, Heymans [/bib_ref] Postvaccination myocarditis, which may result from vaccine-related autoimmunity, has been reported as a rare adverse event after especially smallpox vaccination. [bib_ref] Myopericarditis following smallpox vaccination among vaccinia-naive US military personnel, Halsell [/bib_ref] However, in many cases, the underlying cause has not been identified. The currently available vaccines have been shown to be highly effective against severe coronavirus induced disease 2019 . [bib_ref] Efficacy and safety of the mRNA-1273 SARS-CoV-2 vaccine, Baden [/bib_ref] There are some reports of myopericarditis as a rare complication of mRNA COVID-19 vaccination, especially in adolescent males. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] [bib_ref] Myocarditis temporally associated with COVID-19 vaccination, Rosner [/bib_ref] [bib_ref] Myocarditis after BNT162b2 and mRNA-1273 vaccination, Larson [/bib_ref] Breakthrough infections, such as the Delta or Omicron variant of SARS-CoV-2, have raised great concern about the necessity for repeated dose of the vaccine. [bib_ref] Transmission event of SARS-CoV-2 delta variant reveals multiple vaccine breakthrough infections, Farinholt [/bib_ref] Especially, the Omicron variant, a highly mutated virus of SARS-CoV-2, poses a very high risk of infection and has been designated as a variance of concern by the World Health Organization (WHO) with recommendations for continued vaccination. [bib_ref] Omicron variant of SARS-CoV-2: Genomics, transmissibility, and responses to current COVID-19 vaccines, Araf [/bib_ref] A case of myopericarditis after the second dose of COVID-19 mRNA-1273 vaccine in a 23-year-old man with a prior event of myopericarditis is described. In this case, we focused on the mechanism of myopericarditis after COVID-19 vaccination using mRNA vaccine, and discussed the indication of repeated administration of mRNA vaccine against the risk of myopericarditis.
## Case report
The patient was a 23-year-old Japanese man with a history of myopericarditis 3 years earlier, which resolved after 1 week of hospitalization. Despite the lack of serological evidence, he was tentatively diagnosed as having had viral myopericarditis due to his clinical course, new chest pain, changes in ECG (ST segment elevation at inferior leads), left ventricular wall motion abnormalities in echocardiogram with a trivial inferior pericardial effusion, and transient changes in laboratory results (creatine kinase 473 U/L, creatine kinase-isoenzyme MB 42.3 U/L elevation, etc.). There were no other identifiable causes of the symptoms and findings. This diagnosis corresponded to the possible case of myopericarditis according to the CDC's definition of myopericarditis. [bib_ref] Use of mRNA COVID-19 vaccine after reports of myocarditis among vaccine recipients:..., Gargano [/bib_ref] These abnormalities improved 5 days after admission without medication. Ever since his cardiac function returned to normal, he had not taken any medication. Due to the COVID-19 pandemic, vaccination was made available to and recommended even for healthy young people in Japan. He received his first dose of the COVID-19 mRNA-1273 (Moderna) vaccine without any particular symptoms. Four weeks later, he received the second dose, after which he had a high fever of 38.9°C the next day. At midnight 2 days ABSTRACT: Vaccination is important for the prevention of coronavirus-induced disease 2019 (COVID-19) caused by SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) and to protect persons with a high risk for complications. There have been reports of myopericarditis following COVID-19 vaccination, especially in adolescent males and young adults. Breakthrough infections, such as the Delta or Omicron variant of SARS-CoV-2, have raised great concern about the necessity for repeated doses of the vaccine. A case of myopericarditis after the second dose of COVID-19 mRNA-1273 (Moderna) vaccine in a 23-year-old man with a prior episode of viral myopericarditis is presented. He received the second dose of the COVID-19 mRNA vaccine, after which he developed persistent midsternal chest pain and he was subsequently transferred to our emergency department. An echocardiogram showed a trivial inferior pericardial effusion with diffuse left ventricular systolic dysfunction. He was treated with colchicine from the first day of hospitalization with a diagnosis of myopericarditis. His chest pain had resolved by the third day, and left ventricular wall motion was dramatically improved by the seventh day of hospitalization. A strong response to the second vaccination in the present case suggests that the prior history of myopericarditis is evidence of strong congenital or acquired immunological features in this individual. Individuals with such a strong immune response may be more likely to develop myopericarditis after mRNA vaccination. Immunization against COVID-19 is currently recommended from a risk-benefit standpoint. We advised the patient to avoid additional COVID-19 mRNA vaccines because of this episode. The risk of COVID-19 weighed against myopericarditis associated with the mRNA vaccination should be considered on a case-by-case basis. This case may help us better understand the mechanism of myopericarditis following COVID-19 mRNA vaccination.
KeyWORDS: Myocarditis, pericarditis, myopericarditis, COVID-19, SARS-CoV-2, mRNA vaccine, case report ReCeIVeD: April 18, 2022. ACCePTeD: October 23, 2022.
## Type: case report
# Funding:
The author(s) received no financial support for the research, authorship, and/or publication of this article.
## Declaration of conflicting interests:
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Clinical Medicine Insights: Case Reports following his second vaccination, he developed persistent midsternal chest pain. The pain had not subsided by early morning the following day, and he was subsequently transferred to our emergency department and admitted to our hospital.
The following vital signs were recorded: temperature of 36.9°C, blood pressure of 112/72 mmHg, pulse rate was 86/ min, respiratory rate of 30/min, and oxygen saturation of 100% on room air. Physical examination was unremarkable, with no pericardial friction rub on auscultation. His initial electrocardiogram showed sinus rhythm with atrial premature complexes in a pattern of bigeminy and right axis deviation, ST segment elevations in leads II, aVf, V4-6, and tall T waves in leads V3-6. An echocardiogram showed a trivial inferior pericardial effusion with diffuse left ventricular systolic dysfunction, the ejection fraction was less than 50% (Supplemental Video1-6).
His initial laboratory tests showed a high white blood cell count of 11 000/μL (normal range 3300-8600/μL), neutrophil-dominant (75.7%). The C-reactive protein (CRP) level was 4.29 mg/dL (normal ⩽0.14 mg/dL), high-sensitivity troponin T (TnT) was 0.554 ng/mL (normal ⩽0.1 ng/mL), and B-type natriuretic peptide (BNP) was 68.8 pg/mL (normal ⩽18.4 pg/mL). A SARS-CoV-2 antigen test of specimen obtained from nasal mucosa was negative. Serology of various viruses known as causes of myopericarditis were all negative.
He was treated with colchicine (0.5 mg once daily) from the first day of hospitalization with a presumptive diagnosis of myopericarditis. For follow-up, we performed bedside echocardiography daily until discharge and electrocardiogram was performed several times.
CRP peaked on the second day of hospitalization (6.13 mg/ dL), TnT peaked on the third day (1.34 ng/mL), and BNP peaked on the fourth day (207.8 pg/mL) [fig_ref] Figure 1: Clinical course [/fig_ref].
Contrast-enhanced cardiac magnetic resonance imaging (MRI) performed on day 5 showed linear midmyocardial late gadolinium enhancement in the septal and apical walls of the left ventricle, consistent with acute myocarditis [fig_ref] Figure 2: Contrast-enhanced cardiac magnetic resonance imaging performed on day 5, short-axis views [/fig_ref].
Coronary angiography and endomyocardial biopsy were performed on day 6. Angiography showed normal coronary arteries, and neither obvious lymphocytic infiltration of the myocardium nor cardiomyocyte damage were observed on myocardial biopsy specimens [fig_ref] Figure 3: Endomyocardial biopsy specimens show no obvious lymphocytic infiltration of the myocardium and... [/fig_ref].
His chest pain had resolved by the third day, and left ventricular wall motion was dramatically improved by the seventh day of hospitalization with the ejection fraction of 70%. He was discharged on colchicine and an angiotensin-converting enzyme inhibitor (ACE-I; enalapril 2.5 mg once daily) to prevent subsequent cardiac dysfunction. Colchicine was continued for 1 month after discharge and ACE-I for 6 months. This patient had no recurrence of chest pain for more than 6 months after discharge, and serial echocardiography confirmed the disappearance of the remaining pericardial fluid 1 month after discharge. Although this episode was not fatal and there was no residual cardiac dysfunction, we advised the patient to avoid additional COVID-19 mRNA vaccines because of this episode. were transiently elevated. These findings, including pericardial effusion, improved over the clinical course. BNP transiently increased several days after these changes, and returned to the normal range at discharge. Colchicine was continued for 1 month after discharge and ACE-I for 6 months. WBC, white blood cell (cell count/μL, normal range 3300 -8600/μL); CRP, C-reactive protein (mg/dL, normal ⩽0.14 mg/dL); TnT, high-sensitivity troponin T (ng/mL, normal ⩽0.1 ng/mL); BNP, B-type natriuretic peptide (pg/mL, normal ⩽18.4 pg/mL); CK-MB, creatine kinase-isoenzyme MB (U/L, normal ⩽12 U/L).
## Fujibayashi and kajinami
## 3
# Discussion
A phase 3, randomized, clinical study reportedly showed the safety and efficacy of the mRNA-1273 vaccine describing no safety concerns besides transient local and systemic reactions. [bib_ref] Efficacy and safety of the mRNA-1273 SARS-CoV-2 vaccine, Baden [/bib_ref] However, the CDC recently reported cases of a possible association between the COVID-19 mRNA vaccine and myocarditis/pericarditis.Data from the US Vaccine Adverse Event Reporting System (VAERS) up to June 11, 2021, show reports of 1784 possible cases of myopericarditis after approximately 300 million doses of COVID-19 mRNA vaccine, 67% of which were observed after the second vaccination. Males accounted for 79% of these reports, the majority of whom were under 30 years of age and the median age was 24 years. The median time from vaccination to onset of symptoms was 3 days, and the estimated rate of myopericarditis was 12.6 cases per million doses of the second mRNA vaccine in persons between the ages of 12 and 39 years. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] Most of these patients had elevated levels of troponin T and CRP, none had eosinophilia, and most who underwent magnetic resonance imaging showed late myocardial gadolinium enhancement. Regarding treatment, there are reports of the use of non-steroidal anti-inflammatory drugs (NSAIDs), colchicine, prednisone, etc., and most cases were reported to improve within approximately 1 week. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] [bib_ref] Myocarditis temporally associated with COVID-19 vaccination, Rosner [/bib_ref] [bib_ref] Myocarditis after BNT162b2 and mRNA-1273 vaccination, Larson [/bib_ref] However, since fulminant myocarditis was reported in several cases, careful observation is required. [bib_ref] Myocarditis after BNT162b2 mRNA vaccine against covid-19 in Israel, Mevorach [/bib_ref] [bib_ref] Fulminant myocarditis and systemic hyperinflammation temporally associated with BNT162b2 mRNA COVID-19 vaccination..., Abbate [/bib_ref] In the present case, the course from mRNA vaccination to symptom onset was consistent with these previous reports, and transient elevations of biochemical parameters and left ventricular dysfunction on echocardiogram were observed. In addition, as in previous reports, endomyocardial biopsy showed no obvious myocardial infiltration, 5,6 but magnetic resonance imaging revealed late myocardial gadolinium enhancement. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] [bib_ref] Myocarditis temporally associated with COVID-19 vaccination, Rosner [/bib_ref] [bib_ref] Myocarditis after BNT162b2 mRNA vaccine against covid-19 in Israel, Mevorach [/bib_ref] [bib_ref] Fulminant myocarditis and systemic hyperinflammation temporally associated with BNT162b2 mRNA COVID-19 vaccination..., Abbate [/bib_ref] [bib_ref] Patients with acute myocarditis following mRNA COVID-19 vaccination, Kim [/bib_ref] [bib_ref] In-Depth evaluation of a case of presumed myocarditis after the second dose..., Muthukumar [/bib_ref] Based on these findings, we diagnosed our case as confirmed myopericarditis according to the CDC's definition of myopericarditis. [bib_ref] Use of mRNA COVID-19 vaccine after reports of myocarditis among vaccine recipients:..., Gargano [/bib_ref] The mRNA-1273 vaccine is an mRNA-based vaccine that encodes the full-length spike glycoprotein of SARS-CoV-2. Vaccine mRNA stimulates host cells to synthesize spike glycoproteins and the host immune response produces antibodies against the target virus. There are several possible mechanisms by which this mRNA vaccine may cause myocarditis.
First, a molecular mimicry mechanism between synthesized spike glycoproteins and unknown cardiac proteins has been suggested. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] [bib_ref] Myocarditis after BNT162b2 and mRNA-1273 vaccination, Larson [/bib_ref] [bib_ref] Vaccine-induced autoimmunity: the role of molecular mimicry and immune crossreaction, Segal [/bib_ref] Antibodies to the SARS-CoV-2 spike glycoprotein have been reported to cross-react with human peptide protein sequences. [bib_ref] Potential antigenic cross-reactivity between SARS-CoV-2 and human tissue with a possible link..., Vojdani [/bib_ref] This could explain the high incidence of myocarditis after the second vaccination. However, since cases of myocarditis have been reported even after the first vaccination, multiple mechanisms need to be considered. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] [bib_ref] Myocarditis after BNT162b2 and mRNA-1273 vaccination, Larson [/bib_ref] Another potential mechanism is an immune response or hyperimmunity to the mRNA vaccine. The mRNA molecule is immunogenic and can be attacked before it reaches the target cell. However, the nucleoside-modified mRNA contained in the vaccine has its immunogenicity reduced. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] [bib_ref] Suppression of RNA recognition by Toll-like receptors: the impact of nucleoside modification..., Karikó [/bib_ref] In the presence of a genetic predisposition, the immune response to mRNA does not decrease after nucleoside modification, and the immune system may detect the mRNA in the vaccine as an antigen, leading to a pro-inflammatory cascade and activation of the immune pathway, resulting in myocarditis. In addition, the high prevalence of myocarditis after mRNA vaccination among adolescents may be explained by innate and hyperimmune mechanisms. [bib_ref] Myocarditis with COVID-19 mRNA vaccines, Bozkurt [/bib_ref] [bib_ref] Could sars-coronavirus-2 trigger autoimmune and/or autoinflammatory mechanisms in genetically predisposed subjects?, Caso [/bib_ref] If repeated antigen exposure enhances the immune response, this can also explain why myocarditis has been more prevalent after the second vaccination.
Interestingly, the present case had a history of suspected viral myopericarditis, and this, therefore, plays an important
## 4
Clinical Medicine Insights: Case Reports role in our consideration of the possible mechanisms of myopericarditis after mRNA vaccination. The history of myocarditis, in this case, does not seem to indicate a past coronavirus infection. If past coronavirus infection sensitization had induced an immune response to the vaccine, a stronger immune response would have been observed after the first vaccination. Therefore, the prior history of myopericarditis in the present case is considered to indicate strong congenital or acquired immunological features in the individual. Such individuals may be more likely to develop myopericarditis after mRNA vaccination. However, the limitation of this argument is that it cannot be said that the current event is not a relapse of the initial event.
Kojima et al reported the case of mRNA vaccine-related myocarditis with prior history of myocarditis associated with Campylobacter jejuni. [bib_ref] Report: Myocarditis associated with COVID-19 mRNA vaccination following myocarditis associated with Campylobacter..., Kojima [/bib_ref] In this case, they described that the mechanism of myocarditis induced by mRNA vaccination remains unclear, and that there may be something more in addition to the common pathogenesis mechanism if there is a history of myocarditis. They advised the patient to avoid the booster COVID-19 mRNA vaccine because of the episode.
Vaccination against COVID-19 infection is currently recommended as the benefit outweigh the risk and show a favorable balance for people over 12 years of age. When considering the indications for repeated vaccination, the present case will help us in discussing the pros and cons of vaccination in each case. The risk of COVID-19 weighed against myopericarditis associated with the mRNA vaccination should be considered on a case-by-case basis. Accumulation of cases with a similar course may elucidate the potential mechanism of myopericarditis following mRNA vaccination.
# Author contributions
KF: principle investigator, patient management, manuscript preparation. KK: overall direction of the patient management and manuscript preparation.
# Ethics statement
Ethical review and approval were not required for the study on human participants in accordance with the local legislation and institutional requirements. The patient's informed written consent was obtained; patient confidentiality and data were protected and encrypted.
## Orcid id
Kosuke Fujibayashi https://orcid.org/0000-0001-5567-1424
# Supplemental material
Supplemental material for this article is available online.
[fig] Figure 1: Clinical course. Myocardial enzymes (TnT, CK-MB) and inflammatory markers (WBC, CRP [/fig]
[fig] Figure 2: Contrast-enhanced cardiac magnetic resonance imaging performed on day 5, short-axis views (A), 4-chamber views (B). Linear midmyocardial late gadolinium enhancement is seen in the septal and apical walls of the left ventricle (red arrow). [/fig]
[fig] Figure 3: Endomyocardial biopsy specimens show no obvious lymphocytic infiltration of the myocardium and no cardiomyocyte damage. (Hematoxylin and eosin staining, original magnification ×40, The white bar indicates 200 μm). [/fig]
|
Comparison of nocturnal symptoms in advanced Parkinson’s disease patients with sleep disturbances: pramipexole sustained release versus immediate release formulations
# Introduction
The impact of nonmotor symptoms of Parkinson's disease (PD) on patients' quality of life is comparable to the impact of the characteristic motor symptoms of PD. [bib_ref] Multifactorial sleep disturbance in Parkinson's disease, Albers [/bib_ref] [bib_ref] Health-related quality of life in early Parkinson's disease: the impact of nonmotor..., Duncan [/bib_ref] Sleep disturbances are one of the major nonmotor complaints, and they often manifest even before the clinical onset of PD. [bib_ref] Health-related quality of life in early Parkinson's disease: the impact of nonmotor..., Duncan [/bib_ref] [bib_ref] A timeline for Parkinson's disease, Hawkes [/bib_ref] These sleep problems may worsen with disease progression and deterioration of motor-related abilities. [bib_ref] REM sleep behavior disorder and motor dysfunction in Parkinson's disease -a longitudinal..., Bugalho [/bib_ref] [bib_ref] Hallucinations and sleep disorders in PD, Goetz [/bib_ref] [bib_ref] The relevance of dopaminergic level in nocturnal disability in Parkinson's disease: implications..., Stocchi [/bib_ref] It is estimated that 74%-98% of patients exhibit symptoms of sleep disturbances, including insomnia, abnormal movements during sleep (eg, periodic leg movements and rapid eye movement sleep behavior disorders), and excessive daytime sleepiness. [bib_ref] Multifactorial sleep disturbance in Parkinson's disease, Albers [/bib_ref] [bib_ref] The relevance of dopaminergic level in nocturnal disability in Parkinson's disease: implications..., Stocchi [/bib_ref] However, sleep disturbances are often overlooked in clinical settings and are poorly examined in PD research.
Dopamine has a complex role in the sleep-wake cycle; therefore, several aspects of sleep-related problems are possibly related to dopamine. [bib_ref] Dopamine receptor agonists and sleep disturbances in Parkinson's disease, Chaudhuri [/bib_ref] As a common first-line strategy for treating PD, the use of dopamine receptor agonists such as pramipexole (PPX) may help with the sleep architecture and also improve nocturnal motor symptoms. Research has shown that treatment that is delivered continuously provides more constant stimulation of striatal dopaminergic receptors, which is associated with greater reduction in risk of motor complications and sleep disturbances, than the intermittent, short-acting dopaminergic agents. [bib_ref] The relevance of dopaminergic level in nocturnal disability in Parkinson's disease: implications..., Stocchi [/bib_ref] [bib_ref] The therapeutic concept of continuous dopaminergic stimulation (CDS) in the treatment of..., Stocchi [/bib_ref] PPX, approved in the US and Europe since 1998 for both early and advanced PD, is a nonergot dopamine agonist with selective affinity for dopamine receptor of the D 2 /D 3 subfamily. The early formulation is available as immediate release (IR) tablets and is orally administered three times daily due to the half-life of 8-12 hours. Since 2010, PPX has been available as once-daily sustained release (SR) formulation to meet the needs of patients, clinicians, and caregivers for an effective but simplified treatment regimen. Both PPX SR and IR contain the same active substance and possess identical receptor profile, efficacy, and receptor binding; [bib_ref] Once-daily pramipexole for the treatment of early and advanced idiopathic Parkinson's disease:..., Antonini [/bib_ref] [bib_ref] Pramipexole extended release: a novel treatment option in Parkinson's disease, Eisenreich [/bib_ref] furthermore, the bilayer push-pull osmotic tablet of PPX SR ensures a continuous delivery of PPX at zero-order rate kinetics for an extended period of time, independent of pH and agitational intensity. [bib_ref] Development and evaluation of push-pull based osmotic delivery system for pramipexole, Wakode [/bib_ref] Studies in healthy volunteers have demonstrated that the once-daily PPX SR is bioequivalent to the three-times daily PPX IR, with respect to the area under the plasma concentration-time curve from time 0 to 24 hours (AUC 24 ) and the maximum plasma concentration (C max ). [bib_ref] Pharmacokinetics of a once-daily extended-release formulation of pramipexole in healthy male volunteers:..., Jenner [/bib_ref] Other than the advantage of less frequent dosing, PPX SR is also associated with linear pharmacokinetics and smaller peak-trough fluctuations in healthy volunteers, which suggest potentially better drug tolerability. [bib_ref] Pharmacokinetics of a once-daily extended-release formulation of pramipexole in healthy male volunteers:..., Jenner [/bib_ref] In randomized, double-blind, Phase III studies of early and advanced stage PD, PPX SR demonstrated similar efficacy to PPX IR and significantly greater efficacy than placebo, with or without background levodopa. [bib_ref] Extended-release pramipexole in early Parkinson disease: a 33-week randomized controlled trial, Poewe [/bib_ref] [bib_ref] Extended-release pramipexole in advanced Parkinson disease: a randomized controlled trial, Schapira [/bib_ref] The adverse event (AE) profiles of both the formulations were similar, mostly of mild or moderate intensity, with no unique safety signals observed in PPX SR. [bib_ref] Pramipexole extended-release: a review of its use in patients with Parkinson's disease, Frampton [/bib_ref] These earlier studies did not assess the treatment effects of different formulations on PD-related sleep problems.
In the present study, we retrospectively analyzed nocturnal symptoms of patients receiving PPX SR versus IR from a PPX clinical trial (ClinicalTrials.gov Identifier: NCT00466167) conducted in advanced PD patients with sleep disturbances while on stable levodopa.
# Materials and methods
## Patients and study design
This was a retrospective exploratory analysis of patients with advanced idiopathic PD experiencing motor fluctuations while on levodopa, who took part in the multicenter, double-blind, Phase III clinical trial on the efficacy, safety, and tolerability of PPX SR and IR. 14 The data were collected between May 2007 and November 2008 from 76 sites located across 14 countries. In each country, local institutional review boards and ethics committees approved the clinical trial; details are available from the original publication, and the clinical trial data are all publicly available. [bib_ref] Extended-release pramipexole in advanced Parkinson disease: a randomized controlled trial, Schapira [/bib_ref] The trial was conducted in accordance with the principles of the Declaration of Helsinki, and all patients provided their written informed consent.
The clinical trial included patients who were at least 30 years of age with a diagnosis of idiopathic PD at Hoehn and Yahr stage 2-4 (on time) for at least 2 years before the trial and were on a stable dose of levodopa at least 4 weeks before baseline. Patients were not allowed any dopamine agonists within 4 weeks before the start of the trial. Over 7 weeks after assignment of treatment arms, patients received
## 2019
Pramipexole sr versus ir in PD patients with sleep disturbances a stepwise uptitration of medication according to their ratings on the Patient Global Impression-Improvement scale. Patients then entered a maintenance phase at optimized dosage for the remainder of the study, administered as 0.375, 0.75, 1.5, 2.25, 3.0, 3.75, or 4.5 mg once daily for the SR formulation and 0.125, 0.25, 0.50, 0.75, 1.0, 1.25, or 1.5 mg three times daily for the IR formulation. Patients who took within 80%-120% of correct total dosage (calculated based on dose in mg), as assessed by physical count of returned study medication, were considered as compliant. Further details of this clinical trial have been reported. 14 For the purpose of this retrospective analysis, patients with sleep disturbances at baseline and received either PPX SR or PPX IR (optimized at 0.375-4.5 mg/day) were included from the trial. Patients with sleep disturbances were defined as those who scored ,90 on the PD Sleep Scale (PDSS). [bib_ref] The Parkinson's disease sleep scale: a new instrument for assessing sleep and..., Chaudhuri [/bib_ref] [bib_ref] Clinical assessment of nocturnal disability in Parkinson's disease: the Parkinson's Disease Sleep..., Chaudhuri [/bib_ref] assessments Baseline clinical information of patients was captured using the Unified Parkinson's Disease Rating Scale (UPDRS), the Hoehn and Yahr stage, and patient-rated questionnaires including the PDSS and the Epworth Sleepiness Scale (ESS). [bib_ref] A new method for measuring daytime sleepiness: the Epworth sleepiness scale, Johns [/bib_ref] The main efficacy outcome measures were the adjusted mean changes at week 18 from baseline in PDSS total mean and domain scores. The PDSS is a patient-rated 15-item questionnaire that measures six domains of nocturnal sleep problems in PD, including global quality of night's sleep (questions 1-3, 14), nocturnal restlessness (questions 4 and 5), nocturnal psychosis (questions 6 and 7), nocturia (questions 8 and 9), nocturnal motor symptoms (questions 10-13), and daytime dozing (question 15). [bib_ref] Improvements in nocturnal symptoms with ropinirole prolonged release in patients with advanced..., Chaudhuri [/bib_ref] Patients were asked to rate the severity of their sleep problems on a visual analog scale of 1-10, ranging from the worst score ("awful" or "always" at the left extremity) to the best score ("excellent" or "never" at the right extremity). A PDSS total score of ,90 indicates sleep disturbances (score ranges from 0 to 150); a higher PDSS score reflects better sleep quality.
Overall safety and tolerability were based on AEs that occurred during the study. Since daytime sleepiness is a recognized adverse effect of dopamine agonists, 20 this was assessed by the adjusted mean changes at week 18 from baseline for both ESS total score and PDSS question 15 ("Have you unexpectedly fallen asleep during the day?"). The ESS is a patient-rated eight-item scale that measures a person's general level of daytime sleepiness in different situations commonly encountered in daily life. The chances of dozing off are rated on a scale of 0 (no chance) to 3 (high chance).
A total score of .10 suggests the presence of possible excessive daytime sleepiness.
# Statistical methods
The efficacy analysis included all randomized patients who received PPX SR or PPX IR and observed PDSS ,90 at baseline. To assess the treatment outcomes of PPX SR and IR in terms of efficacy, the adjusted mean changes (ie, PDSS total score and six PDSS domains) at week 18 from baseline were compared using the analysis of covariance on the observed cases, after adjusting for the pooled country and the corresponding baseline scores. The safety analysis included all patients who were treated with PPX. The AEs were summarized descriptively. ESS and PDSS question 15 were analyzed using the same method as the efficacy analysis. All analyses were carried out using the SAS software version 9.4.
# Results
## Baseline and clinical characteristics
A total of 119 PD patients reported sleep disturbances at baseline (PDSS ,90; PPX SR, n=59; PPX IR, n=60). Baseline characteristics were similar between PPX SR and IR; the only notable difference was that the proportion of off-time Hoehn and Yahr stage 4-5 was higher for the PPX IR group, although both the PPX groups reported similar mean scores across the UPDRS sections [fig_ref] Table 1: Baseline demographic and clinical characteristics of PD patients with sleep disturbances Note [/fig_ref].
## Compliance
The proportion of compliant patients was 98.3% by week 18 for both the PPX groups. Only one patient in each group reported a compliance of ,80%.
## Efficacy outcomes
The mean PDSS total scores at baseline were 65.4 and 72.8 in PPX SR and IR groups, respectively [fig_ref] Table 2 ;: Figure 1 [/fig_ref]. Better recovery from sleep disturbances at week 18 from baseline was observed for PPX SR than for PPX IR (adjusted mean changes, 28.5 versus 21.7 points), at a difference magnitude of 6.8 points in the adjusted mean change of PDSS total scores (95% CI, −2.8 to 16.3; P=0.165); however, this difference did not achieve statistical significance ( safety Among all the 119 treated patients included in this retrospective analysis, the incidences of patients with any AE were similar between PPX SR and IR groups (62.7% versus 70.0%; [fig_ref] Table 3: adverse event overall summary [/fig_ref]. PPX SR group reported lower incidence of patients with drug-related AEs than PPX IR group (42.4% versus 55.0%). The numbers of patients who reported severe AEs and serious AEs were comparable between both PPX SR and IR groups. Among patients who were discontinued from treatment due to AEs, five were on PPX SR and one was on PPX IR.
In terms of daytime sleepiness, the adjusted mean ESS total score remained within the normal range of ,10 at week 18 (8.9-9.5 points) for both PPX SR and IR groups [fig_ref] Table 4: Daytime sleepiness from baseline to week 18 in advanced PD patients with... [/fig_ref]. Both the treatment groups did not differ in the magnitude of the adjusted mean change in ESS total scores at week 18 from baseline; however, the magnitude of the PPX SR group was reported to be slightly greater than that of PPX IR group (adjusted mean changes, −0.8 versus −0.1; P=0.457). Similarly, patients receiving PPX SR reported better improvements in PDSS question 15, although the overall difference from PPX IR group was not significant (adjusted mean changes, 3.4 versus 2.9; P=0.4161).
# Discussion
The aim of this retrospective analysis was to assess the treatment effect of PPX SR and IR on nocturnal symptoms among advanced PD patients with sleep disturbances. At the end of the treatment, PPX SR observed improvements of nocturnal symptoms compared with PPX IR, and both the PPX formulations were generally well tolerated.
Earlier studies have demonstrated that both PPX SR and IR have comparable efficacy and safety profiles in patients with PD, including daily living and motor symptoms (as measured by UPDRS part II and part III). The treatment efficacy reported in most clinical studies was sustained by the multiple PPX IR dosages taken throughout the daytime. Since dopamine receptor agonists have been suggested to help and improve sleep quality through the effects on motor function and the reversal of nocturnal off-time related symptoms, a continuous dopaminergic stimulation (CDS) may mean an overall better sleep maintenance for patients with PD. [bib_ref] Multifactorial sleep disturbance in Parkinson's disease, Albers [/bib_ref] [bib_ref] Dopamine receptor agonists and sleep disturbances in Parkinson's disease, Chaudhuri [/bib_ref] [bib_ref] Nocturnal symptom complex in PD and its management, Chaudhuri [/bib_ref] In rat models, early morning akinesia was better prevented by a continuous delivery of PPX via subcutaneously implanted minipumps compared with PPX via injections; the positive outcome was attributed to the availability of more CDS. [bib_ref] Continuous dopaminergic stimulation by pramipexole is effective to treat early morning akinesia..., Ferger [/bib_ref] Previous single-arm or placebo-controlled clinical trials such as the SP826, [bib_ref] Rotigotine transdermal system for control of early morning motor impairment and sleep..., Giladi [/bib_ref] EASE-PD, [bib_ref] Ropinirole 24-hour prolonged release: randomized, controlled study in advanced Parkinson disease, Pahwa [/bib_ref] and the RECOVER [bib_ref] Rotigotine effects on early morning motor function and sleep in Parkinson's disease:..., Trenkwalder [/bib_ref] have shown benefits in the management of sleep disturbances in PD with sustained release nonergot dopamine agonist treatment with or without levodopa. To date, only one randomized,
## 2021
Pramipexole sr versus ir in PD patients with sleep disturbances parallel-group study has compared the sleep outcomes of PD patients taking different release formulations of nonergot dopamine agonist; no statistical difference in adjusted mean change in PDSS total score at week 24 from baseline was observed between ropinirole prolonged release versus ropinirole IR in patients with PD (5.7 versus 5.8 points; P=0.937). [bib_ref] PREPARED: Comparison of prolonged and immediate release ropinirole in advanced Parkinson's disease, Stocchi [/bib_ref] It is noted that the ropinirole study consisted of patients with generally normal sleep, which explains the smaller magnitude of sleep improvement after treatment compared with the current PPX study.
In the current retrospective analysis, an improvement in sleep was observed in PD patients with existing sleep disturbances for both PPX formulations. Although the PPX SR-IR difference of 6.8 points (adjusted mean changes in PDSS total score, SR=28.5 versus IR=21.7 points, P=0.16) did not meet statistical significance, this change does not necessarily preclude any clinically meaningful benefit to patients, and therefore, the latter remains to be evaluated. We noted that for studies in PD patients using PDSS second version (PDSS-2), 27 an improvement of .3.44 points or deterioration of .2.07 points on the PDSS-2 total score is recommended as the threshold for minimal clinically important difference in nocturnal sleep quality. [bib_ref] Minimal clinically important difference on Parkinson's Disease Sleep Scale 2nd version, Horváth [/bib_ref] Consistently, we observed that the patients receiving PPX SR in this study reported a slightly greater proportion of recovery from sleep disturbances at week 18 compared with those on PPX IR (PDSS total scores of $90, SR=55.1% versus IR=50.8%). Further research comparing PPX SR with PPX IR and possible intervention trials including PDSS-2 are needed to help address these clinical gaps in PD patients with existing sleep disturbances.
The retrospective analyses confirmed the results of previous randomized controlled trials demonstrating that PPX SR and IR are equally safe and well tolerated by patients with advanced PD. [bib_ref] Extended-release pramipexole in advanced Parkinson disease: a randomized controlled trial, Schapira [/bib_ref] Both patient groups reported similar improvements in daytime sleepiness after 18 weeks of receiving PPX regardless of formulation, as measured by both ESS and PDSS question 15.
PD is a chronic disease that requires long-term treatment management; however, there is a significant concern of treatment nonadherence among patients with PD, varying between 10% and 67%, which leads to poor response to therapy. [bib_ref] Medication adherence in patients with Parkinson's disease, Malek [/bib_ref] Once-daily treatment regimens have been shown to significantly improve treatment adherence in chronic diseases compared with twice-daily or thrice-daily regimens, potentially increasing adherence days by up to 44%. [bib_ref] Effect of medication dosing frequency on adherence in chronic diseases, Saini [/bib_ref] In another community survey, 83.1% of patients have ranked the need for a reduction in medication frequency as a high priority in their expectation toward treatment (rating of 1-3 on the 6-point scale). [bib_ref] Requirements for Parkinson's disease pharmacotherapy from the patients' perspective: a questionnaire-based survey, Wüllner [/bib_ref] In terms of symptoms alleviation, 76.8% ranked sleep disturbance as top three and 55.4% ranked early morning akinesia as "very important". According to these studies, a once-daily treatment regimen would contribute toward adherence to treatment, medication dosage, and timing to ensure that patients receive the available care. Considering the existing pool of patients who are already on PPX IR treatment, the success of switching from IR to SR formulations was evidenced by improvements in UPDRS Parts II+III in 86% of patients with advanced PD within 1 week after the switch. [bib_ref] Success rate, efficacy, and safety/tolerability of overnight switching from immediate-to extendedrelease pramipexole..., Schapira [/bib_ref] Longer follow-up durations of switches to PPX SR from IR reported equal or better outcomes in the adherence and motor symptoms of patients with advanced PD, without severe adverse effects. [bib_ref] Adherence to antiparkinson medication in a multicenter European study, Grosset [/bib_ref] [bib_ref] Efficacy and safety of a once-daily extended-release formulation of pramipexole switched from..., Takanashi [/bib_ref] The limitations of this analysis are inherent in the retrospective and exploratory design. The sample size of the study was not planned to confirm the efficacy outcomes, and only the actual observed data were used for analysis without any further imputations for dropouts (eg, last observation carried forward). Nonetheless, the positive and consistent sleep outcomes observed with PPX SR over PPX IR are encouraging and may serve as a pilot for prospectively designed studies to collect further supporting evidence. In conclusion, in advanced PD patients with sleep disturbances and on stable levodopa, both the PPX formulations showed improvements in sleep disturbances. However, the current observations on the numerical difference between groups suggest that future prospectively designed research is necessary to confirm the effects of PPX SR for improved sleep.
[fig] Figure 1: ir, P-value (95% ci) 0.6, P=0.5020 (−1.1 to 2.2) Notes: a adjusted = least-square means adjusted for pooled country and baseline scores; b positive change implies improvement. Data are presented as mean ± sD. Abbreviations: ir, immediate release; PD, Parkinson's disease; PDss, Parkinson's Disease sleep scale; PPX, pramipexole; sr, sustained release. advanced PD patients with sleep disturbances (PDss ,90): adjusted mean changes at week 18 from baseline in PDss total and domain scores. Note: error bars show the standard error. Abbreviations: ir, immediate release; PD, Parkinson's disease; PDss, Parkinson's Disease sleep scale; PPX, pramipexole; sr, sustained release. [/fig]
[table] Table 2 ;: Figure 1). At week18, 55.1% of patients receiving PPX SR and 50.8% on PPX IR reported PDSS total scores of $90, that is, demonstrated improvements in sleep. For each of the six individual PDSS domains, patients receiving PPX SR reported numerically greater adjusted mean change at week 18 from baseline than those receiving PPX Drug Design, Development and Therapy 2018:12 submit your manuscript | www.dovepress.comThe final mean (SD) PPX dosages at week 18, after flexible uptitration, were 2.9 (1.5) and 2.9 (1.4) mg/day for PPX SR and PPX IR groups, respectively. [/table]
[table] Table 1: Baseline demographic and clinical characteristics of PD patients with sleep disturbances Note: a The sample size was n=42 for PPX sr and n=56 for PPX ir. [/table]
[table] Table 2: PDss total and domain scores from baseline to week 18 of advanced PD patients with sleep disturbances [/table]
[table] Table 3: adverse event overall summary [/table]
[table] Table 4: Daytime sleepiness from baseline to week 18 in advanced PD patients with sleep disturbances [/table]
|
Inequalities in birth weight and maternal education: a time-series study from 1996 to 2013 in Brazil
Maternal education represents one of the most important social determinants of inequality in birthweight (BW) in developing countries. The present study sought to investigate secular trends in health inequality considering the difference in mean BW between extremes of maternal educational attainment in Brazil. Using a time-series design, data from 6,452,551 live births which occurred in all Brazilian state capitals from 1996 to 2013 were obtained from the Information System on Live Births. Secular trends of the difference in mean birth weight between low (<8 years of schooling) and high (≥12 years of schooling) educational attainment were analyzed. The main finding was that differences in mean birth weight between the two extremes of maternal educational attainment decreased over time. There was a significant decrease in mean BW in neonates born to mothers with higher educational attainment, and a slight increase in those born to mothers with lower educational attainment. One of the key factors involved in decreasing inequality was an increase in the number of antenatal visits. In view of these results, we conclude, that despite a slight increase of mean birth weight among mothers with low education, the reduction of inequality in pregnancy outcomes over time in Brazil is attributable to a worsening scenario for mothers who are better off rather than to improvements for the most vulnerable group of mothers.
stress are more likely to engage in high-risk behaviors such as smoking, alcohol intake, and psychoactive drug use, and are more likely to be overweight or obese-all factors which have been associated with lower birth weight. Maternal education is also among the most relevant social determinants of child health; mothers with low educational attainment are at higher risk of delivering low birthweight infants.
Average birth weight has increased in many countries over the last three decades, due to improved access to antenatal care and population-wide improvements in socioeconomic condition. However, some recent studies have found that mean birth weight has declined in some developed countries, although specific causes have not been identified. Therefore, mean birth weight has presented different secular trends among different countries, under the influence of several determinants.
Brazil has one of the highest rates of low birth weight among developing countries, at approximately 8,0% and remaining stable over the last 10 years. This rate is strongly influenced by sociodemographic factors and by access to adequate health care, and thus reflects high inequality among the different social classes and geographic regions of the country. However, the assessment of rates of low birth weight alone is not able to show the small impacts of the various individual social determinants of birth weight over time.
Therefore, the objective of this study is to evaluate the influence of inequality between extremes of maternal education on birth weight, considering the trend of absolute and relative differences in mean birth weight over time among mothers with higher and lower educational attainment in an 18-year time series in Brazil.
# Methods
## Data source and study population. this time-series study included all live births which occurred in 26
Brazilian state capitals and the Federal District from 1996 to 2013. Data were obtained from the Information System on Live Births (Sistema de Informação sobre Nascidos Vivos, SINASC), which covers 96% of all live births in Brazil. This open-access database is available through the Ministry of Health website (http://datasus.saude. gov.br) or directly on the Health Information page of the Unified Health System Department of Informatics -DATASUS (http://datasus.saude.gov.br/informacoes-de-saude/tabnet).
Multiple births were excluded, as were newborns weighing <500 g or >8,000 g. These exclusions represented 6.02% of all live births during the period of interest . The decision to use only information about births occurring in Brazilian capitals was based on the recognition that, in these cities, epidemiological surveillance services are more consolidated and have a more robust structure, which ensure greater completeness and reliability of data.
Overall, there were 6,452,551 live births in Brazilian state capitals and the Federal district during the study period, representing 11.92% of all live births in the country. Assessment for completeness showed a missing data rate <0.5% Outcome, variable, and covariables of interest. Trends in mean birth weight (in grams) over time, presented as a continuous variable, were the outcome of interest. The investigated variable, maternal educational attainment, was dichotomized into two extremes: low and high (<8 and ≥12 years of formal schooling, respectively). Thus, extreme levels of maternal education were used because we understand that these categories are more subject to the potential impact of access to health services and technologies during antenatal and perinatal care on birth weight.
The following covariables were considered: maternal age (10-17, 18-34, ≥35 years), number of previous live births (0 or ≥1), number of antenatal visits (none, <6, or ≥7 visits), gestational age (˂37 or ≥37 weeks), and mode of delivery (vaginal or cesarean). State capitals were grouped by the five geographic regions of Statistical analysis. Birth weight means were calculated yearly for each Brazilian capital. The absolute and relative differences in mean birth weight between the two extremes of maternal education were calculated to assess the inequality between them.
The annual means of birth weight differences between the two extremes of maternal education was estimated using a linear mixed modelStatistical analysis was performed in PASW Statistics, Version 18.0 (SPSS Inc., Chicago, IL), with a 5% significance level and 95% confidence intervals (CIs).
# Results
During the study period, there was a significant reduction in adolescent pregnancies. There was also an increase in cesarean sections among mothers with low educational attainment, which, by the end of the period, was equivalent to approximately half of the percentage observed among mothers with high educational attainment. The descriptive variables used in the study, including maternal socioeconomic conditions, number of previous live births, number of antenatal visits, percentage of pregnancies in adolescence, cesarean sections, and preterm deliveries, are shown in.
The findings point out a gradual decrease in the frequency of women with low educational attainment, from 55.4% in 1996 to 18.5% in 2013. Accordingly, there was an increase in the frequency of women with high educational attainment, from 7.2% in 1996 to 24.4% in 2013.
The absolute difference in mean birth weight among infants born to mothers with high vs. low educational attainment was 90.22 g in 1996, gradually decreasing to 5.05 g in 2011-the smallest difference for the whole period of interest. The relative difference also decreased, from 2.87% in 1996 to 0.41% in 2013. www.nature.com/scientificreports www.nature.com/scientificreports/ presents the means of birth weight differences of neonates born to mothers with high vs. low educational attainment, adjusted for covariables, in a sequential linear mixed model in the two different periods of analysis.
## Year of birth
In both periods, the number of antenatal visits (Model d) reduced the inequality between birth weight means in the two extremes of maternal education. Maternal age (Model b) had a negative effect, mainly in the second period (2004-2013), increasing the inequality in means of birth weight between the two extremes of maternal education. The covariables number of children, gestational age, and mode of delivery did not show any effect on the differences in birth weight means between the two groups of educational attainment during the study period.
# Discussion
The results of this study showed a significant decline in mean of birth weight among infants born to mothers with a higher education level, leading to a significant reduction of general mean of birth weight during the time series. In terms of inequality in means of birth weight according to maternal education, we present a scenario compatible with the hypothesis of similarities in the inequality proposed for us previously, in which intense use of technologies and care coexisting with lack of access to them can lead a similar outcome 9,27,28 . Besides, the general increaseshows the secular trends and annual mean birthweight of neonates born to mothers with high and low educational attainment in Brazilian state capitals from 1996 to 2013. In this figure, the curves of the mean weight of newborns born to mothers with high ("Mean_high") and low ("Mean_low") education are presented, as well as the secular trends of average weight of newborns born to mothers with discharge ("Jp high") and low ("Jp low") education, assessed through Joinpoint Regression. Each of these curves has different outlines and colors marked each year. It is possible to observe that the mean birth weight among neonates born to mothers with high educational attainment declined steadily throughout the study period, while remaining stable among infants born to mothers with low schooling, with only a slight increase from 2005 onward. www.nature.com/scientificreports www.nature.com/scientificreports/ in maternal education could contributed to amplify the severity of inequality between the two extremes of social, even though we noticed a narrowing trend of the outcome. In this case, the less educated mothers represent the most vulnerable social group in the country leading to an scarce access to health care and educational policies during the study period.
In addition, several maternal and child health indicators-including number of antenatal visits, multiparity, and the teenage pregnancy rate-showed overall improvement during the study period. In contrast, there was an increase in primiparous women aged 35 or older, which has been associated with the increase in cesarean sections and premature births in .
Other factors could be playing a significant role in this scenario, such as improvements in income distribution, with a general reduction in poverty leading to less maternal malnutrition, better quality of health care, and increased access to health care, especially for underprivileged social groups. Besides, social programs, such as conditional cash programs, showed significant impact on infant health leading to reduction of infant mortality in less privileged communities.
However, these positive factors did not produce a detectable effect on birth weight for women in unfavorable socioeconomic conditions, possibly indicating that this group has experienced extreme social exclusion and marginalization. Thus, considering this scenario, the stability of mean birth weight in the less educated group can be considered a net positive.
The effect of the number of antenatal visits on birth weight could be explained in two different directions according to educational level. For more educated mothers, antenatal care can be associated with intensive use of new technologies, such as in vitro fertilization, cesarean section, and ultrasound, leading to possible reductions in mean birth weight even without an increase in preterm birth rates. In mothers with lower schooling, expanded access to antenatal care can lead to better care of basic health needs, contributing to a reduction in acute illness and increased vaccination and nutritional follow-up rates.
Antenatal care has long been considered an important factor related to birth weight. Inadequate numbers of antenatal visits are associated with low birth weight, while increased access to antenatal care is associated with improved birth weight and other favorable pregnancy outcomes. Mothers who attended fewer than 6 visits had a relative risk for low birth weight of 2.47 compared to those who attended 6 or more visits, even after adjustments for income and maternal education. Six or more antenatal visits were associated with better outcomes for pregnant women with hypertensive disorders, while fewer visits or no antenatal care whatsoever were related to stillbirth and perinatal death, as well as higher risk of urinary tract infections, chorioamnionitis, and streptococcal vaginitis.
The steady increase in rates of cesarean section presented influence in reduction in the difference in mean birth weight in both study periods, but more so in the second period proportionally. These increases occurred in women at both extremes of maternal education, but in those with lower educational attainment. The greater proportional increase in CS rates in this less privileged group of women could represent significant improvement in access to adequate health care, instead of excess of intervention, amplifying fetal survival, initially for heavy newborns. Brazil has one the highest rates of cesarean section in the world (52%), which has been associated with high rates of preterm birth. This practice may also be associated with changes in the concept of fetal viability and interventions to prevent fetal deaths.
Preterm births also have played a role in the birth weight trends observed herein, especially in the group of mothers with higher educational attainment. A study that evaluated preterm births between 2000 and 2010 in the most developed areas of Brazil suggested that successful interventions increased the rate of preterm live births by 45.2%. Likewise, changes in the profile of pregnant women in the country, with an increase in the number of mothers aged 35 or older, have increased the risk of pre-existing health conditions and complications during pregnancy, such as hypertensive disorders. This factor could be related to the decrease in mean birth weight observed, and has been previously described as the "low birth weight paradox", whereby better socioeconomic development is associated with a higher rate of low birth weight. Limitations of this study include the lack of information about preexisting maternal clinical background, obstetric diseases, and intercurrent events during pregnancy. Data on pre-gestational body mass index, smoking, alcohol intake, psychoactive drug use, and maternal nutrition were not available. Inaccurate recording of gestational age in weeks and the high rate of incompleteness of the variable race/skin color made the inclusion of these variables in the study impossible.
On the other hand, the national scope of this study, the number of live births included in the analyses (more than 6 million) and its representativeness within the Brazilian population, the 18-year period of analysis (from 1996 to 2013), and the acknowledgment of maternal obstetric characteristics associated with birth weight are among the strengths of this study.
These results reflect the demographic, epidemiological, and perinatal transitions in the country over the last two decades, which have had an impact on mean birth weight. Similar trends have been reported in other countries, such as the United States and China, with decreases in birth weight in the last year. Social advances and improvements in maternal education and maternal and child care have failed to exert a positive impact on birth weight, probably due to increased medical and technological interventions at all maternal education levels. Besides, the limited change in mean birth weight observed among infants born to women in unfavorable socioeconomic conditions may suggest that advances have not reached the most vulnerable strata of society.
In conclusion, despite a slight increase of mean birth weight among mothers with low education, this study indicates that the reduction of social inequalities in birth weight in Brazil came at the expense of worse outcomes in the better-off group, as opposed to improvements in mothers with low educational attainment. These findings suggest that the current epidemiological profile of birth outcomes in Brazil is characterized by the presence of "equality in inequality", whereby similar outcomes occur as a result of differential effects of interventions and different social scenarios, leading to a merely apparent reduction of social inequality between extreme social strata.
Finally, the results found allow us to affirm that the reduction of inequality in pregnancy outcomes over time in Brazil is mostly attributable to a worsening scenario for mothers who are better off rather than to improvements for the most vulnerable group of mothers. |
Attitudes and values among the Swedish general public to using human embryonic stem cells for medical treatment
Background: The use of human embryonic stem cells (ES cells) for the development of medical therapies is surrounded with moral concerns. The aim of this study was to assess the public's attitudes toward the use of ES cells for treatment of Parkinson's disease (PD) and other diseases, what factors are most important to consider when using ES cells for drug development, and if there is an association between religious beliefs and attitudes toward using ES cells for medical treatment.Methods:A randomly selected sample of the Swedish public, aged 18-87-years-old, completed an online survey (n = 467). The survey assessed socio-demographics, religious views, perceived moral status of the embryo, and attitudes toward using ES cells for medical treatment of PD and other diseases. Adjusted odds ratios (ORs) and 95% confidence intervals (CI) for positive vs. negative attitude toward using ES cells for drug development were computed using logistic regression.Results:The respondents were positive about using ES for treatment; specifically, 70% totally agreed that it is acceptable to use ES cells for treatment of PD, while 40% totally agreed that it is acceptable to use ES cells for treatment but induced pluripotent cells is just as efficient. Religion being of little importance in one's life was associated with a positive attitude toward using ES cells for treatment of PD (adjusted OR 6.39, 95% CI 2.78-14.71). The importance of being able "to access new, effective treatments against diseases that do not have any treatment available" was ranked as the most important factor to consider when using ES cells for drug development.Conclusion:Most respondents are positive about using ES cells for drug development, and making effective treatments accessible to those who do not have any. However, these attitudes are influenced by the specific disorder that the drug development is intended for, as well as the religious views and perceived moral status of the early embryo.
# Introduction
Stem cells are currently being used to develop therapies for a number of incurable diseases [bib_ref] The future of stem cell therapies for Parkinson disease, Parmar [/bib_ref]. Human embryonic stem cells (ES cells) are immature, pluripotent cells that have the ability to divide an infinite number of times and develop into any specialized cell. Generation of ES cells involves extracting cells from the blastocyst on the fifth day after conception. The embryos used to generate ES cells are often surplus embryos donated by couples undergoing in vitro fertilization (IVF) treatment. Couples undergoing IVF in Sweden can choose either to donate to other couples for reproductive purposes, to donate the embryos for research purposes, or to discard them. In Sweden, it is not allowed to use specifically created embryos for the purpose of research. Induced pluripotent cells (iPS cells) are created from reprogrammed specialized cells [bib_ref] Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures..., Takahashi [/bib_ref] [bib_ref] Induction of pluripotent stem cells from adult human fibroblasts by defined factors, Takahashi [/bib_ref]. Researchers hope that both ES cells and iPS cells can be used in medical treatment in the field of regenerative medicine to replace the cells that have been damaged or have died. Parkinson's disease (PD) is one of the conditions, where the development of treatment has come the furthest [bib_ref] Ethical and safety issues of stem cell-based therapy, Volarevic [/bib_ref]. Parkinson's disease is a chronic progressive disorder that can decrease an individual's independence and quality of life. The etiology of PD is still unknown, and there is currently no cure for PD. Instead, treatment aims to reduce symptoms and maintain a quality of life using medication and nonpharmacological interventions. Therefore, development of effective disease modifying treatments is important in order to help patients and their families. The potential and limitations of ES-derived and iPS-derived cell products in the treatment of PD is summarized in a recent review by Parmar et al. [bib_ref] The future of stem cell therapies for Parkinson disease, Parmar [/bib_ref] There are currently three ongoing clinical trials, two with a cell product produced from ES cells and one with cells produced from iPS cells. A case report of one patient transplanted with an autologous iPS cell-derived product is also published [bib_ref] Personalized iPSC-derived dopamine progenitor cells for Parkinson's disease, Schweitzer [/bib_ref].
Unlike, iPS cells, the use of ES cells is surrounded by certain moral concerns because early embryos are discarded during production. The view of the embryo's moral status varies, depending on the outlook on life and has previously been the subject of extensive discussions. Strong voices in society, from activists and religious leaders, have opposed the use of ES cells [bib_ref] Applying stem cell therapy in intractable diseases: a narrative review of decades..., Brianna [/bib_ref]. These moral objections have influenced legislation, including restricting the time for usage to 14 days in order to protect the special status of the unborn [bib_ref] The limited application of stem cells in medicine: a review, Poulos [/bib_ref] , and have in some countries led to banning research with ES cells [bib_ref] Ethical and safety issues of stem cell-based therapy, Volarevic [/bib_ref]. Furthermore, manufacturing medical treatment involves commercialization, which risks embryos being considered as commodities, with a lack of respect for human life. When the time comes for treatment with cell products derived from ES cells or iPS cells to be introduced on the market, policymakers or relevant qualified regulatory agencies need to decide whether commercialization of these cells should be allowed for treatment of patients, or under what conditions they should be allowed. Before such decisions are made, it is crucial that different publics' views are considered, which is empathized in the Swedish Patent Act (1967:837, 2020:541) Chapter 1, Article 1c, laying down that "a patent is not granted for an invention where the commercial exploitation would be contrary to public order or morality".
Much of the previous research has focused on the IVF couples' views. A systematic review from 2014 found that couples who donated their embryos to research perceived it as a better alternative than the destruction of embryos and as an opportunity to help others. Couples assigning embryos a high moral status were less likely to donate their embryos for research purposes [bib_ref] Factors associated with the donation and non-donation of embryos for research: a..., Samorinha [/bib_ref]. In a Swedish cross-sectional study on the attitudes toward embryo donation, about half of the respondents were positive toward donating for research purposes [bib_ref] Attitudes towards embryo donation in Swedish women and men of reproductive age, Wånggren [/bib_ref]. In a Chinese study, 59% of the IVF couples preferred to discard surplus embryos rather than donate them to research, stating a lack of information and distrust in science as significant reasons for their decision [bib_ref] Patients' attitudes towards the surplus frozen embryos in China, Jin [/bib_ref]. A cross-sectional study assessing if Latin American IVF couples would donate embryos found that the attitudes varied depending on the specific purpose of donation. The respondents were most positive about donation for treatment with embryonic stem cells (73.6%), while 63.8% would donate to heterosexual couples for reproductive purposes, and 45.1% to single women [bib_ref] Embryo donation among Latin-Americans who have attended assisted reproduction techniques: a first..., Álvarez-Díaz [/bib_ref] , indicating that the purpose of donation matters for people's attitudes. The source of the embryo, either being surplus of created for research purpose, also matters for peoples attitude for usage [bib_ref] Human embryonic stem cell research: ethical views of Buddhist, Hindu and Catholic..., Sivaraman [/bib_ref]. In a qualitative study from 2022 with members of the Swedish general population, the participants had a positive attitude toward using embryonic stem cells to treat PD; however, they raised several conditions under which the treatment should be allowed [bib_ref] Perceptions on using surplus embryos for the treatment of Parkinson's disease among..., Grauman [/bib_ref]. Some found early embryo donation to be completely unproblematic, while others had ethical concerns connected to their perception of the early embryo, in terms of being a "potential life. " The participants who viewed the early embryo as a potential life were still positive about donating, since they considered the benefit of helping PD patients find an efficient treatment as being more important than the negative aspects related to using the early embryos. However, some thought that ES cells would no longer be used if iPS cells or other treatments were as efficient. This emphasizes the importance of specifying the disease and the presence of available treatments when assessing the public's attitudes toward using ES cells for treatment. In this study, we wanted to follow-up on the findings from the previous Swedish interview study to assess the prevalence of the attitudes and values in the Swedish population regarding the use of ES cells for drug development. The study's aim was to study the general public attitude about the use of ES cells for medical treatment in general and of PD in particular, and what factors are most important to them when considering using ES cells for drug development. Further, we aimed to investigate if there is an association between the importance of religion in one's life and attitudes toward using ES cells for medical treatment in general and of PD in particular.
# Methods
## Data collection and study population
This was a cross-sectional study among a random sample of Swedish citizens, aged 18 or older (n = 2755), drawn from the Swedish State Personal Address Register. The register includes Swedish residents. An invitation letter included information about the study, together with a link to an online survey. Thirty-four letters were not delivered. Data collection continued until we reached a minimum of 400 respondents. Of the total 2721 who received the invitation, 467 participated in the study (17%). Fourteen people informed us that they did not want to participate, five were not able to participate because of their health status, two did not have a computer, and one did not understand Swedish. Data were collected between March and May 2021.
## Research ethics
The study was approved from the Swedish Ethical Review Authority before the data collection started (Dnr 2019-06539). Informed consent was obtained from all the respondents. Data were presented in such way that no individual can be identified. All the methods were carried out in accordance with relevant national and international guidelines and regulations.
## The survey and variables
The online survey included questions about age, gender, country of birth (Sweden or other), occupation, highest level of education completed (later dichotomized as 1 = university, 0 = all other), pharmaceutical use, and health literacy (Sufficient or limited) [bib_ref] The communicative and critical health literacy scale-Swedish version, Wångdahl [/bib_ref].
After responding to these questions, the respondents were asked to read a text informing them about human embryonic stem cells and how they can be used to treat PD (Additional file 1).
Thereafter, the respondents were asked to rank eight statements regarding factors to consider when using surplus embryos for drug development based on importance . The statements were formulated based on interests expressed by patients, the general public, and couples having cryopreserved embryos in previous qualitative interviews.
Attitudes toward using leftover embryos for drug development were assessed by three items and Likert scales ranging from 1 to 5 (1: fully agree, 2: highly agree, 3: partially agree, 4: highly disagree, and 5: fully disagree). The point of departure was a scenario where leftover early embryos from IVF treatment were donated by the couple for the specific purpose, without financial compensation. The statements the respondents were to decide upon were: It is acceptable that surplus embryos from fertility treatments are (1) used for treatment of Parkinson's disease, (2) used for treatments of other types of diseases, and (3) used for treatment of diseases, even if induced pluripotent cells may be used with similar results. In the analyses, the ratings from the Likert scales were dichotomized into positive attitude (1-2 point) and negative attitude (3-5 point).
The respondents' perception of the moral status of a couple of days old human embryo was assessed using the question, "The human is perceived to have a special moral position, in the sense of having rights just by being human. What moral position does a human embryo that is only a few days old have? Respondents could choose between four statements (The embryo is just a lump of cells = 1; it is meaningless to discuss its moral status / the embryo has a moral status that is in between being just a lump of cells and being a human being = 2; The embryo in its moral status is closer to being a human than just a lump of cells = 3); and the embryo has the same moral status as a human being = 4). The variable was dichotomized into low status (1-2 points) and high status (3-4 points).
The importance of religion in the respondents' life was assessed on a 5-point Likert scale (1: very little importance, 2: fairly little importance, 3: neither great nor little importance, 4: quite great importance, and 5: great importance). The variable was dichotomized into little importance (1-3 point) and high importance (4-5 points).
## Statistical analyses
Descriptive statistics are presented with mean and standard deviation for continuous variables and as frequencies for categorical variables. The difference in attitudes toward using surplus embryos for drug treatment between the three purposes of use was tested with Wilcoxon Signed Ranks Test. The Mann-Whitney U test and Spearman's correlation test were further used to test for differences in attitudes based on age, sex, education, country of birth, health literacy, moral status of the embryo, and religion. The associations between religion and attitudes toward using surplus embryos for drug treatment were estimated as odds ratios (ORs) with 95% confidence intervals (CI). Three separate binary logistic regression analyses were made, one for each attitude item. A negative attitude was the reference category. The crude model included religious views. Model 1 included variables in model 1 and age, sex, education, health literacy, country of birth, and use of pharmaceuticals. Model 2 included variables from the previous models, and moral status of the embryo. All analyses were performed using SPSS 25.
# Results
In this cross-sectional survey study, the age of the respondents ranged from 18 to 87 years and encompassed about an equal share of men and women. Almost half of the respondents had a university degree, and 86% were born in Sweden [fig_ref] Table 1: Characteristics of the respondents [/fig_ref].
About half of the respondents stated that religion had a very little meaning in their life, while only 4.4% stated that religion had very much meaning in their life. A majority of the respondents perceived the early embryo as a lump of cells, while 3.9% thought that the early embryo has the same moral status as a human being [fig_ref] Table 2: Descriptive statistics of meaning of religion and perceived moral status of the... [/fig_ref].
## The public's attitude toward using surplus embryos for development of medical treatment
The respondents were overall very positive toward using surplus embryos for drug development ; a low median and mean indicate a positive attitude). There was a statistically significant difference in attitude, based on the specific conditions of use. The respondents were most positive toward using surplus embryos for treating Parkinson's disease specifically; second, most positive toward using surplus embryos; and least positive toward using surplus embryos, when iPS is just as efficient. The Mann-Whitney U test further showed that respondents who were less religious, male, younger, born in Sweden, and had sufficient health literacy were more positive toward using embryos. Respondents who perceived the embryo as a human or closer to a human than a lump of cells were most negative toward using embryos for drug development.
The associations of religious view and attitudes toward using surplus embryos were estimated in three separate logistic regression models for each attitude statement, shown in . Respondents who stated that religion had little importance in their life were more likely to have a positive attitude toward using surplus embryos compared to respondents with religion playing an important role in their lives. The association was strongest when surplus embryos were used for treating PD (adjusted OR 6.39, 95% CI 2.78-14.71), followed by treatment of other diseases (adjusted OR 3.47, 95% CI 1.56-7.71), and using embryos despite iPS cells being just as efficient (adjusted OR 1.75, 95% CI 0.92-3.34).
The complete regression models are presented in the Additional file 2, indicating that respondents that perceive the embryo as a lump of cells or between a lump of cells and human were more likely to be positive toward using surplus embryos for treating PD (7.86 (3.43-18.02)). Neither age, sex, health literacy nor country of birth were associated with attitudes. Additional analysis found that respondents who stated that religion had little importance in their life were more likely to perceive the moral status of the embryo as lower (adjusted OR 4.61, 95% CI 2.27-9.36), Additional file 2: .
## Most important factors to consider when using es cells for drug development
The respondents were asked to rank statements of aspects to consider when using surplus embryos for drug development based on importance . To have access to new, effective treatments against diseases
## Perceived moral status of embryo
The embryo is just a lump of cells; it's pointless to discuss its moral standing 252 (58.1)
The embryo has a moral position that lies in the middle of being just one lump of cells and being a human being 128 (29.5)
The moral position of an embryo is closer to being a human being than just a lump of cells 37 (8.5)
The embryo has the same moral status as a human being 17 (3.9) Descriptive statistics of respondents' (n = 435) attitudes toward using leftover embryos from IVF treatment for treatment, donated by the couple, without financial compensation for the purpose of drug development
## Do not agree at all n (%)
It is acceptable to use leftover embryos for treatment of Parkinson's disease The association of importance of religion in life with attitudes toward using embryos for treatment of I) Parkinson's disease II) of other diseases III) of diseases, although iPS would be as efficient
Associations are presented using odds ratio (OR) and 95% confidence intervals (CI), and with respondents finding religion to be of fairly high or very high importance, as the reference groups (ref )
OR odds ratio, CI confidence interval. that do not have any, was considered the superior most important thing to consider. This was followed by the need to reduce the risk of severe side-effects, and the need for transparency from medical companies involved in the drug development.
# Discussion
The respondents who answered the survey were generally very positive about using ES cells for treatment, especially when it comes to treatment of PD. They were a bit less positive concerning the treatment of other diseases. However, if iPS cells are available with equal efficacy, the respondents were less positive about using ES cells. This corresponds to the findings reported in our previous interview study with individuals from the Swedish population [bib_ref] Perceptions on using surplus embryos for the treatment of Parkinson's disease among..., Grauman [/bib_ref]. It is possible that individuals with a special interest in PD were more likely to respond to the survey, thus resulting in a bias in the sample. That could explain them being more positive toward accepting treatment for PD in particular. Another explanation is that the attitudes were biased by the information about PD at the beginning of the survey, making the respondents feel more strongly about PD. However, finding an effective treatment for diseases that lack an available treatment was ranked the most important aspect to consider when utilizing ES cells for treatment purposes, dominating all other factors considered. This can explain why treatment for PD was ranked higher than treatment for other diseases, since current PD therapies only address symptoms. The results were not surprising as it corresponds to a previous qualitative study, where the respondents did not distinguish between PD and other serious chronic diseases lacking a curable treatment [bib_ref] Perceptions on using surplus embryos for the treatment of Parkinson's disease among..., Grauman [/bib_ref]. However, the respondents had a low ranking for the importance of using iPS cells instead of ES cells to produce medical products for moral reasons, indicating that, although it was considered important, other aspects are more important. However, there were 25 respondents (5.7%) who ranked this as the most important aspect; these respondents were probably more religious and perceived the moral status of the early embryo as a human or close to a human. In this study, we found an association between attitude toward using ES cells for medical treatment purposes and religious views and the perceived moral status of the early embryo among the Swedish general population. A Swiss study found that IVF couples' perception of the moral status of an embryo and their stated religious views were independently predictive of their attitudes toward donation of surplus embryos for various purposes [bib_ref] Attitudes of couples towards the destination of surplus embryos: results among couples..., Mohler-Kuo [/bib_ref]. The number of people in our study who stated that religion was important or very important in their life, and the number of respondents who viewed the moral status of the embryo as a human or close to a human, were relatively few (11.4% and 12.6%, respectively). According to the Culture Map, provided by the World Value Survey, Sweden is a country with the most secular values in the world. The culture map also shows that the values of migrants from e.g., Afghanistan, Iraq, Somalia, living in Sweden differ substantially compared with native Swedes. Sharma et al. found that Asian immigrants in the United States were less likely to donate surplus embryos for research use [bib_ref] Asian immigrants to the United States are less likely to donate cryopreserved..., Sharma [/bib_ref]. In our study, about 14% of the respondents were born outside of Sweden, but we do not know from which countries.
The assessment of religious concerns in this study may be seen as a limitation, and, indeed more nuanced views may have been revealed if we had used an instrument and questions from life view research that captured religious Statements reflecting what the public thinks is important to consider when using surplus embryos for drug development
Respondents were asked to rank the statements from most important (1) to least important. Descriptive statistics are presented using medians, inter-quartile ranges (IQR), and the number of times the statement was ranked as the most important (n)
## Statements median (iqr) ranked as most important (n)
It is important to access new, effective treatments against diseases that do not have any 1 (2) 255
It is important to try to reduce the risks for severe side-effects associated with medical treatments 2 (2) 23
It is important with transparency and scrutiny if medical companies are involved in developing treatment with embryonic stem cells
## (2) 32
It is important that legislators consider the different values in society when deciding on the use of embryos for medical treatment
## (3) 45
Patients should be informed on how their treatment has been developed before deciding whether they accept treatment or not
## (4) 17
If we allow embryos to be used for medical treatment, there is a risk that we, in the future, will allow unwanted usage 6 (3) 14
For moral reasons, it is preferable to use induced pluripotent stem cells instead of embryonic stem cells 6 (3) 25
Couples donating embryos should not be compensated economically for their donation 6 (4) 25
beliefs, morally salient values, as well as fundamental attitudes as a basic component of a religious worldview [bib_ref] Religion and understanding, Jeffner [/bib_ref]. However, the aim was to see if the respondents selfreported as seeing religion as something important in their lives, whether their views were influenced by beliefs about the world and human beings, their morally salient values, or basic attitudes related to if they had a more positive or a more negative view of life. The instrument used did, in fact, reflect both on beliefs and values, e.g., the biological and moral status of a human embryo. Furthermore, previous studies have assessed religious views in similar ways, also finding that individuals with moderate to strong religious beliefs were less likely to donate to research [bib_ref] Patient attitudes to donation of embryos for research in Western Australia, Burton [/bib_ref]. Further empirical explorations in life-and worldview research may provide a richer understanding of how both comprehensive and fragmented religious views may direct moral positions. It may also be noted that from other religious contexts, the views are both aligned with and different from the majority view in this study. For instance, Sivaraman and Noor have studied ethical views about using ES cell among religious leaders from different religions in Malaysia. They found that the Islamic and Hindu leaders allowed the use of surplus embryos, to which they attached less moral weight, while they prohibited the use of "research embryos. " Furthermore, Islamic leaders have concluded that the use of ES cells is allowed for the purpose of maintaining health, and mandatory if it can relieve people's suffering. The Buddhist leaders in their study did not see any ethical difference between the use of either source of embryos, approving the use of ES cells as long as it is done with caution and care. The Catholic leaders, however, opposed the use of embryos, whether surplus or research embryos. The different standpoints reflected different views on the moral value of the embryo, and whether the benefit is more important than the harm. They also found different reflections from leaders from the same religion, expect the Catholic leaders, who repeated the same message as formulated by the Vatican [bib_ref] Human embryonic stem cell research: ethical views of Buddhist, Hindu and Catholic..., Sivaraman [/bib_ref]. Besides religious views and the moral status of the early embryo, other studies have found that men, in general, are positive about allowing embryo donation for research [bib_ref] Attitudes towards embryo donation in Swedish women and men of reproductive age, Wånggren [/bib_ref] [bib_ref] Attitudes toward embryo donation among staff at in vitro fertilization clinics, Wånggren [/bib_ref]. Likewise, the univariate analyses in this study showed that men held a more positive attitude. However, this association was not found in the adjusted analyses. Furthermore, health professionals have been more positive than the public in previous studies on donating embryos for research [bib_ref] Attitudes toward embryo donation among staff at in vitro fertilization clinics, Wånggren [/bib_ref]. Being scientifically oriented and having trust in experts and the medical system are also factors related to a more positive attitude [bib_ref] Factors associated with the donation and non-donation of embryos for research: a..., Samorinha [/bib_ref] [bib_ref] Perceptions on using surplus embryos for the treatment of Parkinson's disease among..., Grauman [/bib_ref]. A limitation of the study is that we have not assessed the respondents' trust in the healthcare systems or whether the respondents worked within the healthcare systems themselves. Neither does the questionnaire assess the importance of considering genetic privacy issues for donors or issues regarding commercialization. Another limitation is the low response rate (17%), restricting the generalization of the results to the Swedish population at large. Our study sample consisted of more individuals with a university degree compared with the Swedish general population. However, a random sample of the general population was invited to participate, and the final sample included an equal number of men and women of various ages and educational-and occupational backgrounds, which is a strength of the study. Therefore, although migrant groups may be missing, it is possible to make a statement about the attitudes and values of native Swedes of various ages.
The literature on the ethics of using stem cells give witness to moral concerns and sometimes to intense discussions [bib_ref] The bioethics of stem cell research and therapy, Hyun [/bib_ref]. Some areas where this is evident are gene therapy, pre-implantation genetic diagnosis, whole genome sequencing, or gene editing. They have all, like stem cell-based research, stirred intense ethical discussions when they first were presented in scholarly journals and reported in public media. Some early research applications with these technologies were indeed premature and should have awaited better evidence but, after some progress and more scientific evidence about benefits and risks, most of them will belong to mainstream medical science. Gene therapy is an example of a promising new technology developed 40 years ago. It met quite some resistance, not the least from religious representatives. Some warned against Gene therapy as a way of "Playing God". Forty years later, there are approved treatments and several clinical trials with gene therapy ongoing. The technology is now moving into the mainstream medical science governed by ordinary regulatory frameworks for clinical trials, despite the fact that in the beginning it was viewed by many as being morally unacceptable. It is against this background that the results of this study are of particular interest, showing that the majority of the general public care foremost about issues relating to medical needs and being able to improve medical treatment. Furthermore, the history of stem cell research have shown us that strong moral values och opinions from some voices in society, e.g., activists or religious leaders, can influence legislation and stop development of valuable treatments for patients. For democratic reasons, it is crucial to make an effort to listen to the voice of the general public as direct stakeholders, and not only those with openly voiced opinions. It is not apparent that policy should always do as the majority thinks, but policy makers should always listen to their perspective, as part of a transparent dialogue to maintain trust for researchers, the health care system and the pharmaceutical industry. It is our hope that policy makers use the findings from this study for deliberation and communication.
# Conclusion
The majority of the respondents are positive towards using ES cells for drug development, and value making effective treatments accessible for those who do not have any. However, these attitudes are influenced by what disorder the drug development is intended for, and religious views and the perceived moral status of the early embryo.
[table] Table 1: Characteristics of the respondents (n = 467) [/table]
[table] Table 2: Descriptive statistics of meaning of religion and perceived moral status of the embryo (n = 467) [/table]
|
SYT-SSX fusion genes and prognosis in synovial sarcoma
A case series of 64 synovial sarcomas was characterized for the SYT-SSX fusion transcripts and statistically analysed in order to correlate molecular data with prognosis and morphology. SYT-SSX1 fusion transcript appeared to be an independent, though not reaching statistical significance (P = 0.183), prognostic factor clearly associated with a reduced metastasis-free survival. Regarding the association between transcript type and histologic subtype we found, a borderline P value (P = 0.067) between the SYT-SSX1 transcript and the biphasic subtype which, subsequently expanding the analysis to 70 cases, turned out to be significant. However, we could not confirm the prediction value of the biphasic subtype for the presence of the SYT-SSX1 transcript since in our hands 6 out 33 (18%) biphasic tumours carried the SYT-SSX2 transcript.
Synovial Sarcoma (SS) is a distinct and fairly common kind of soft tissue sarcoma, occurring most often in young adults and having a predilection for the extremities. Two histologic subtypes are recognized: the biphasic, the most typical but less common sub-type, containing both an epithelial and a spindle cell component; and the monophasic subtype, entirely made up of spindle cells. A rare form of monophasic SS has been described, containing only epithelial-like cells [bib_ref] Differential diagnosis of small round cell tumors, Meis-Kindblom [/bib_ref].
Cytogenetically, SS is characterized by the translocation t(X;18)(p11.2;q11.2) in more than 95% of the cases. This translocation fuses the SYT gene from chromosome 18 with one of three closely related genes SSX1, SSX2 or SSX4, located at the X chromosome. The fused genes form a chimeric protein that seems to deregulate either the transcription or the expression of specific target genes. The current evidence suggests that this non-random SS translocation may be associated to the transactivation of a cascade of genes such as c-MET [bib_ref] Expression of hepatocyte growth factor (HGF)/Scatter factor and its receptor c-Met correlates..., Oda [/bib_ref] , c-KIT [bib_ref] ) c-KIT and c-KIT ligand (SCF) in synovial sarcoma (SS): a mRNA..., Tamborini [/bib_ref] and their ligands, and BCL-2 [bib_ref] Analysis of SYT-SSX fusion transcript and bcl2 expression and phosphorylation status in..., Mancuso [/bib_ref] , yielding two different autocrine loops, one acting on the epithelial differentiation (c-MET), and the other working in an apoptosis-protecting program (BCL-2, c-KIT).
Synovial sarcoma has been traditionally regarded as a highly malignant tumour, but its prognosis varies greatly depending on a number of patients evaluated or clinico-pathologic features taken into consideration. One is tumour size, that has been shown to be an independent predictor of both distant recurrence and mortality [bib_ref] Synovial Sarcoma: a multivariate analysis of prognostic factors in 112 patients with..., Lewis [/bib_ref]. Another is tumour grade. Recently, two grading systems have been proposed for SS. One is morphologybased and depends on the presence of atypia, necrosis and mitotic activity . The second is clinico-pathologic oriented, and takes into account patient's age, tumour size, and the presence of a poorly differentiated tumour component. Both systems have been shown to split SSs into favourable and unfavourable cases [bib_ref] Synovial Sarcoma. Identification of low grade and high risk groups, Bergh [/bib_ref]. Additional biological markers of possible prognostic significance have been recently investigated, including Ki-67 [bib_ref] Association of SYT-SSX fusion types with proliferative activity and prognosis in Synovial..., Inagaki [/bib_ref] , c-myc [bib_ref] Prognostic significance of nuclear accumulation of c-myc and mdm2 proteins in Synovial..., Shen [/bib_ref] , and the co-expression of HGF and c-MET [bib_ref] Expression of hepatocyte growth factor (HGF)/Scatter factor and its receptor c-Met correlates..., Oda [/bib_ref].
Two recent studies have focused on the relevance of the different fusion gene types in affecting the outcome of disease [bib_ref] SYT-SSX gene fusion as a determinant of morphology and prognosis in synovial..., Kawai [/bib_ref] [bib_ref] The SYT-SSX1 variant of synovial sarcoma is associated with a high rate..., Nilsson [/bib_ref]. Both studies showed a significant reduction of metastasis-free survival in patients carrying the SYT-SSXI fusion type of transcript when compared with those carrying the SYT-SSX2 fusion type. Furthermore, SYT-SSX1 and SYT-SSX2 fusion transcripts were shown to be significantly associated with biphasic and monophasic SSs, respectively.
Here we report on our own series of SS patients characterized for the presence and type of SYT-SSX fusion transcripts with the purpose of further verifying the value of the different SYT-SSX gene fusions as determinants of both prognosis and morphology in SS.
# Materials and methods
A total of 72 consecutive cases of SS collected between 1989 and 1999 in which frozen tissue samples were available were retrieved. The criteria applied for the diagnosis of biphasic and monophasic type were those reported byi.e., the presence of epithelial and spindle cell components in varying proportions in the former and a fibrosarcoma-like spindle cell component with no demonstrable epithelial component in the latter. Six cases were excluded from the study for the following reasons: four patients already had metastatic disease at the time of diagnosis, and therefore they could not be assessed for time to occurrence of distant metastases (the primary study endpoint); one
## Syt-ssx fusion genes and prognosis in synovial sarcoma
patient did not receive radical surgery; and another patient had previously a Ewing sarcoma/pPNET of the thoracic region, carrying the t(11;22)(q24;q12) translocation.
The tissue samples analysed molecularly were from primary tumours in 26 patients, from recurrences in 18, and from distant metastases in 20 cases.
The treatment of the patients affected by localized tumours (primary in 26 and recurrent in 18) consisted of conservative surgery in 41, (associated to radiation therapy whenever indicated), and major disarticulation in the other 3. Systemic chemotherapy was administered according to standard protocols, with 31 patients treated by adjuvant or neoadjuvant therapy.
All patients with distant metastases to lung (20 cases) had at least one excision of the pulmonary mass(es), and 9 of them underwent systemic chemotherapy, as well as 22 of the remaining patients.
# Molecular analysis
A frozen tumour fragment from each case was mechanically disaggregated and total RNA was isolated using the RNAzoIB extraction system (TEL-TEST, Friendswood, Texas). One µg of total RNA was reverse-transcribed into cDNA using oligo(dt) primers and reverse transcriptase (Superscript, Gibco BRL, Paisley, UK), according to the manufacturer's recommendations.
The integrity of cDNA was tested by the amplification of the ubiquitous gene β-actine [bib_ref] Determination of hexokinase isoenzyme I and II composition by RT-PCR: Increased hexokinase..., Adams [/bib_ref]. The detection of the putative SYT-SSX1 and SYT-SSX2 fusion gene was carried out with the primers SYT 5′CAA CAG CAA GAT GCA TAC CA3′, SSX1 5′GGT GCA GTT GTT TCC CAT CG3′ and SSX2 5′GGC ACA GCT CTT TCC CAT CA3′ [bib_ref] SYT-SSX gene fusion as a determinant of morphology and prognosis in synovial..., Kawai [/bib_ref]. PCR conditions were the following: 35 cycles of denaturation at 94˚C for 30 sec, annealing at 58˚C for 1 min, and elongation at 72˚C for 1 min. The detection of the putative SYT-SSX4 fusion transcript was carried out by a nested RT-PCR with the following primers: SYT external 5′CAA CAG CAA GAT GCA TAC CA3′ and SSX external 5′TGC TAT GCA CCT GAT GAC GA3′ in the first step and SYT internal 5′AGA CCA ACA CAG CCT GGA CCA3′ and SSX4 5′GGC ACA GCT GTT TCC CAT CA3′ in the second step . The thermal profile for both reactions was identical to that mentioned for the detection of SYT-SSX1 and SYT-SSX2, except that the annealing temperature in the first step was 52˚C. Each reaction included cDNA from normal mesenchymal tissue as negative control. The amplification products of all the PCR reactions were analysed on 2% agarose gel that can clearly detect both the classic SYT-SSX fusion trancripts and all the described variants [bib_ref] Molecular diagnosis of Synovial Sarcoma Characterization of a variant SYT-SSX2 fusion transcript, Fligman [/bib_ref] [bib_ref] Characterization of a variant SYT-SSX1 Synovial Sarcoma fusion transcript, Safar [/bib_ref] All the sequence reactions were carried out using an automated sequencing system (377 DNA Sequencer, ABI PRISM PE, Applied Biosystem) following standard protocols.
# Statistical methods
The association of SYT-SSX gene transcripts (SYT-SSX1 and SYT-SSX2) with the various clinical and pathologic parameters investigated was tested using the Pearson chi-square test, with exact P computation .
The main endpoint of the study was metastasis-free survival. Time to distant metastasis was computed from the date of disease diagnosis to the date of event occurrence, or the date of last follow-up visit for patients free from distant metastases.
Metastasis-free survival curves were obtained with the Kaplan-Meier method. The prognostic value of SSX gene transcripts on metastasis-free survival was assessed using the Cox model, both without and with adjustment for the following covariates: age (< 25 versus ≥ 25 years); tumour site (head and trunk versus limbs) tumour size (> 5 versus ≤ 5 cm); and tumour histologic subtype, (monophasic versus biphasic). The choice of covariates and their categorization was done before the analysis and relied on clinical expertise and literature data. Patients for which the source of tumour samples was a recurrent lesion represent a selected population, in that they entered the study cohort given that they had developed such a recurrence. Their behaviour was expected to differ from that of patients with samples from the primary tumour. Exploration of our data clearly showed that subjects with samples from distant metastases differed prognostically from the remaining ones. For this reason Cox model analyses were stratified according to the source of the tumour sample as follows: local sample cohort, encompassing primary tumour and/or local recurrences; and metastatic sample cohort. The choice of covariates and their categorization was done before the analysis and relied on clinical expertise and literature data.
Patients who had the assessment performed on both types of tissue (6 cases) were included in the local sample cohort. In all such cases, the SYT-SSX gene translocation results yielded always concordant results in the different samples.
The covariates were entered into the Cox model by means of 0-1 indicator variables. We also tested the interaction terms between SYT-SSX gene transcripts and the type of tumour sample analysed or the histologic subtype. The proportional hazard assumption implied by the model was checked using log-minus log survival plots. Hazard ratio (HR) estimates obtained with the Cox model were reported together with the corresponding 95% confidence limits and associated P-value. For a given covariate, a hazard ratio equal to one denotes a lack of prognostic effect, whereas values above (below) one denote an increased (decreased) risk of occurrence of the event in a particular category, compared to the reference category.
Analyses were carried out using SAS™ (SAS Institute Inc., 1990). P-values reported are two-sided; a P-value below 0.05 was considered significant, whereas a P-value between 0.10 and 0.05 was considered borderline.
# Results
# Molecular analysis
Among the 72 patients with SS (42 monophasic and 30 biphasic) analysed for the SYT-SSX fusion genes, the SYT-SSX1 and the SYT-SSX2 fusion types were detected in 44 (61.1%) and 26 (36.1%) of the tumours, respectively. The remaining 2 (2.8%) cases had an SYT-SSX4 fusion transcript. 20 of the 42 (47.6%) monophasic tumours showed a SYT-SSX1 fusion type, 20 (47.6%) a SYT-SSX2, and 2 (4.8%) a SYT-SSX4. Conversely, 24 (80%) of the 30 biphasic SSs showed a SYT-SSX1, and 6 (20%) a SYT-SSX2 fusion type.
By amplifying the DNA across the breakpoint region no electrophoretically different bands were detected. The PCR product of two cases were sequenced and used as positive control for SYT-SSX1 and SYT-SSX2 fusion transcripts, respectively. The PCR products of the two SYT-SSX4 cases were also sequenced, as well as those of six biphasic SS showing an SYT-SSX2 fusion transcript.
# Statistical analysis
The two cases carrying the SYT-SSX4 fusion transcript were discarded. Since six cases had already been eliminated (see Patients and Methods), the statistical analysis was performed on 64 cases, 40 SYT-SSX1 and 24 SYT-SSX2, diagnosed between 1989 and 1999, all of which had loco-regional disease at presentation.
The prevalence of the SYT-SSX1 transcript was higher among males then females, in tumours ≤ 5 cm in size, and in biphasic lesions. Tumour site did not appear to substantially affect the prevalence of the SYT-SSX1 transcript, and there was no evident trend with patient age. A borderline P was obtained for histologic subtype (P = 0.067).
Forty-five distant metastases were recorded. Metastasis-free survival curves according to the type of transcript were substantially overlapping in local sample cohort . On average, metastasis-free survival was about 50% at 5 years. In the metastatic sample cohort, the survival tended to be lower for SYT-SSX1 carrying tumours .
The unadjusted hazard ratio estimate for SYT-SSX (SYT-SSX1 versus SYT-SSX2) was 1.35 (95% confidence limits: 0.71-2.57; P = 0.365). Results of the multivariable Cox regression model are reported in [fig_ref] Table 2: Results of the multivariate Cox model [/fig_ref]. The hazard ratio estimate adjusted for age, tumour site, tumour size and histologic sub-type was 1.61 (0.80-3.23; P = 0.183) overall. The figures obtained according to the type of tissue sample analysed or the histologic subtype were: 1.37 (0.55-3.38) for the local sample cohort, 1.95 (0.71-5.37) for the metastatic sample cohort (P = 0.595 for the interaction between SYT-SSX and type of sample analysed); 1.36 (0.62-3.01) for monophasic tumours, 2.59 (0.67-10.0) for biphasic tumours (P = 0.406 for the interaction between SYT-SSX and histologic subtype). Regarding the remaining covariates, a less favourable prognosis (hazard ratio > 1) was associated with age < 25 years, head/trunk tumour location, tumour size > 5 cm, and monophasic histologic subtype. However, none of these differences was significant on statistical analysis.
# Discussion
The present study was undertaken to further evaluate the predictive value of the two alternative forms of the SYT-SSX fusion transcript (SYT-SSX1 and SYT-SSX2), concerning the prognosis and the morphology of SS.
The statistically valuable case material consisted of 64 cases where samples were obtained from 44 local tumours (26 primary and 18 local recurrent tumours) and 20 metastases. The tumours were classified as 38 monophasic and 26 biphasic and characterized as 40 with SYT-SSX1 and 24 with SYT-SSX2. PCR amplification and the subsequent electrophoretical analysis of the obtained products revealed band size compatible with the presence of the usual SYT-SSX genes fusion types in all the cases. For this reason after having verified the sequence of control cases we did not perform the DNA sequencing of all the PCR products, in the light also of Nilsson et al data which did not find uncommon rearrangements in fusion transcripts of the same size [bib_ref] The SYT-SSX1 variant of synovial sarcoma is associated with a high rate..., Nilsson [/bib_ref].
## Syt-ssx and prognosis in synovial sarcoma 1537
British Journal of Cancer According to the same authors a tumour gene profile assessed on samples from patients enrolled at the time of primary/recurrent tumours may differ from that performed on samples from patients with metastatic disease [bib_ref] The SYT-SSX1 variant of synovial sarcoma is associated with a high rate..., Nilsson [/bib_ref]. By contrast, the translocation (X;18)(p11.2;q11.2) is a non-random genetic hallmark stable over the disease progression. Thus, after a preliminary analysis performed in our series showing that cases with primary tumour or local recurrence samples (local sample cohort) had similar outcome in terms of metastasis-free survival, we split our case material into local and metastatic sample cohorts. In such a way we accounted for the heterogeneity in the tumour material used, at variance with previous work done by [bib_ref] SYT-SSX gene fusion as a determinant of morphology and prognosis in synovial..., Kawai [/bib_ref]. There was also some evidence of a possible imbalance in the distribution of a known prognostic characteristic such as tumour size between SYT-SSX1 and SYT-SSX2 cases . Cox model multivariable analysis was aimed at removing possible confounding due to such an association.
As estimated by hazard ratios (HR) of the multivariate Cox regression model, tumour size turned out to be the factor more strongly associated with poor prognosis (HR = 1.89), followed by the type of SYT-SSX fusion (HR = 1.61). When investigating the effect of the different SYT-SSX transcripts according to the type of sample and histologic subtype, we observed a higher HR in the group of the metastatic sample cohort (HR 1.95) than in the local sample cohort (HR 1.36), and in biphasic tumours (HR 1.95) than in monophasic ones (HR 1.36). All together, the data were consistent with a clear, though not significant, less favourable prognostic effect of the SYT-SSX1 fusion type in metastatic sample cohort and biphasic tumours.
As compared with previous findings our HR estimates were clearly lower. In particular the findings observed by multivariate analysis were 3.0 (95% confidence limits, 1.1-0.80, P = 0.003) in the first series published [bib_ref] SYT-SSX gene fusion as a determinant of morphology and prognosis in synovial..., Kawai [/bib_ref] and 7.4 (1.5-36.0, P = 0.004) in the second one [bib_ref] The SYT-SSX1 variant of synovial sarcoma is associated with a high rate..., Nilsson [/bib_ref]. Notably, the latter considered only patients with SYT-SSX fusion transcripts assessed on primary tumours, whereas in the former, the fusion transcript assessment was performed not only on primary tumour, but also on local recurrence and metastasis.
Regarding tumour size, our results confirm also the wellestablished [bib_ref] Tendosynovial sarcoma: clinicopathologic features, treatment and prognosis, Brodsky [/bib_ref] and recently-reconfirmed evidence [bib_ref] Synovial Sarcoma: a multivariate analysis of prognostic factors in 112 patients with..., Lewis [/bib_ref] [bib_ref] Synovial Sarcoma of the extremities. A clonicopathologic study of 34 cases, icluding..., Machen [/bib_ref] of an association between large tumour size (> 5 cm) and poor clinical outcome. Among the remaining variables considered, the age < 25 years, the axial tumour location and the monophasic pattern correlated, though not significantly, with a less favourable outcome in keeping again with the literature [bib_ref] Synovial Sarcoma. Identification of low grade and high risk groups, Bergh [/bib_ref].
The analysis of association between fusion transcripts and clinico-pathological variables showed, in addition to a higher prevalence of SYT-SSX1 in males, and in tumour smaller than 5 cm in diameter, an association, with a borderline P value (P = 0.067), between such a fusion transcript and the biphasic SS subtype. We observed in fact six (23%) biphasic SSs carrying a sequence-verified SYT-SSX2 fusion transcript. Interestingly three out of six belonged to children, and involved uncommon sites. It is worth mentioning that the Fisher's exact test performed on the entire series of 70 cases (see Materials and Methods), yielded a significant result (P = 0.010) , in keeping with recent reports [bib_ref] SYT-SSX gene fusion as a determinant of morphology and prognosis in synovial..., Kawai [/bib_ref] [bib_ref] Strong association of SYT-SSX fusion type and morphologic epithelial differentiation in synovial..., Antonescu [/bib_ref].
In conclusion we reconfirm the association between the type of SYT-SSX fusion transcript and histologic sub-type. Regarding the clinical outcome, our data show a trend, not reaching statistical significance, of the SYT-SSX fusion transcript as an independent prognostic factor coherent with previous findings. However, in consideration of the heterogeneous effects estimated for this factor in currently available studies, we think that the prognostic effect of SYT-SSX fusion transcript deserves further confirmation on larger patient groups.
## Acknowledgement
This work was supported by grants from 'Ministero della Sanità. Ricerca Finalizzata 1998-ICS030. 1/RF98.32-Italy' and by AIRC (Associazione Italiana per la Ricerca sul Cancro), grant n. 420.198.122. The authors are grateful to Dr Juan Rosai for the critical revision of the manuscript.
[fig] Figure 1, Figure 2: Metastasis-free survival curves according to the type of transcript in the local sample Metastasis-free survival curves according to the type of transcript in the metastatic sample cohort [/fig]
[table] Table 2: Results of the multivariate Cox model: hazard ratio estimates (HR), corresponding 95% confidence limits (CL) and P-value [/table]
|
Foreign Body Induced Osteomyelitis in the Hand – Commonly Missed Clinical and Radiological Diagnosis
[bib_ref] Foreign body granuloma: A diagnosis not to forget, El Bouchti [/bib_ref] [bib_ref] Injuries from palm tree thorn simulating tumoral or pseudotumoral bone lesions, Curiel [/bib_ref]
# Discussion
In CT, wood appears as a linear cylindrical area of increased attenuation, which is better seen in wide window settings [bib_ref] Wooden foreign bodies: Imaging appearance, Peterson [/bib_ref]. It Thorn is an organic foreign body, which can develop foreign body granuloma, resulting in periosteal reaction or even osteolysis. It can also cause synovitis due to a chemical reaction within the joint. If it remains in the sub-periosteal region for some time, it can cause a sub-periosteal reaction with fluid collection followed by periosteal elevation. Subsequently, this causes devitalization of the cortical bone, hence, resulting in a sequestrum [bib_ref] Orthopaedic manifestations of date thorn injury: Case series, Suresh [/bib_ref] [bib_ref] Foreign-body osteitis of the metacarpal bone, Peters [/bib_ref]. The periosteal reaction can mimic conditions like osteomyelitis, bone cysts, or any aggressive bony neoplasms. In soft tissues, the retained plant thorn can also cause foreign body cyst, bursitis, tenosynovitis, and synovitis [bib_ref] Injuries from palm tree thorn simulating tumoral or pseudotumoral bone lesions, Curiel [/bib_ref] [bib_ref] The missed thorn, Barry [/bib_ref]. Since they are not radio-opaque, X-rays are negative for the organic foreign matter until and unless bony lesions develop which warrants further investigations [bib_ref] Orthopaedic manifestations of date thorn injury: Case series, Suresh [/bib_ref]. On ultrasonography, a thorn is highly echogenic due to the presence of entrapped air with acoustic shadowing and can therefore help identify the wooden foreign body [bib_ref] Wooden foreign bodies: Imaging appearance, Peterson [/bib_ref]. may also demonstrate hyper dense lesion, which is suggestive of sequestrum, and indicates the possibility of osteomyelitis [bib_ref] Injuries from palm tree thorn simulating tumoral or pseudotumoral bone lesions, Curiel [/bib_ref]. With MR imaging, these foreign bodies show a variable signal intensity that is equal to or less than that of skeletal muscle on both T1-and T2-weighted images with surrounding inflammatory response [bib_ref] Wooden foreign bodies: Imaging appearance, Peterson [/bib_ref]. Ipaktchi et al. [bib_ref] Retained palmar foreign body presenting as a late hand infection: Proposed diagnostic..., Ipaktchi [/bib_ref] , have suggested a diagnostic algorithm for the detection of these types of radiolucent foreign bodies where high-frequency ultrasonography is preferred first. If only it cannot visualize the lesion, CT/MRI can be used for further evaluation. According to their algorithm, if there is a history of trauma with metallic object then plain radiography should be done and if radiography is negative for a foreign body or if there is history of trauma with non-metallic object, then highfrequency linear ultrasonography should be done in the emergency department (ED). If the ultrasonography is negative, then CT/MRI should be considered. In cases where radiography or ultrasonography is positive, then superficial palpable foreign body should be explored in ED and operating room evaluation should be done for all other foreign bodies. Intraoperative pre-procedural ultrasonography localization of the foreign body should also be done.
# Conclusion
For ultrasound-guided foreign body removal, only a small incision in the skin is required, which must be wide enough for the foreign body to pass through. Surgical forceps are inserted to reach the foreign body under ultrasonography guidance to remove it [bib_ref] US-guided localization and removal of soft-tissue foreign bodies, Cura [/bib_ref]. According to this study, the effectiveness of USguided percutaneous removal can be near 100%. For foreign bodies that are deep and not accessible through ultrasound, a three-dimensional CT data set can be used with intraoperative navigation systems in order to facilitate surgical removal [bib_ref] Effect of navigation system on removal of foreign bodies in head and..., Ji [/bib_ref]. Culture sensitivity of tissue/pus obtained along with the foreign body is usually inconclusive although the osteomyelitis is primarily induced by it [bib_ref] Neglected thorn injury mimicking soft tissue mass in a child: A case..., Gupta [/bib_ref].
In few cases of foreign body-induced osteomyelitis, the patients may not recall a history of foreign body prick. These patients are, therefore, evaluated with CT/MRI as first line study, which sometimes fail in identification of these foreign bodies from sequestrum. Hence, clinicians and radiologists should be highly suspicious for the foreign body as a possible etiology, in patients unresponsive to routine treatment protocols. Radiography for radiopaque, ultrasonography for recent/superficially located and CT are useful for radiodense and wooden foreign bodies with MRI remaining a problem-solving device in few isolated cases.
[fig] Figure 1: (a, b) CT coronal bone window and (c) CT axial soft tissue window. CT scan in bone and soft tissue window setting shows an irregular lytic lesion in the shaft of 5th metacarpal along with irregular cortical thickening with sclerosis. Two areas of a cortical breach are identified. Two hyperdensities seen in the soft tissue medial and anteromedial to the lytic lesion (arrows) were initially thought to be sequestrum. Retrospectively, it turned out to be the organic foreign body (Thorn). [/fig]
[fig] Figure 2: (a, b, c) MRI axial PDFS and (d) MRI sagittal PDFS. Intramedullary altered marrow signals are seen in the 5th metacarpal from the base to the distal physis, appearing hyperintense on PD fatsaturated sequences. There is associated irregular cortical thickening which is breached at multiple places (arrows). There is an extension of abnormal soft tissue from the medullary cavity into the overlying soft tissues through these cortical breaches. Thin linear hypointense lesion (arrows) seen anteromedial to the cortical breaches in (a, c) was retrospectively assessed to be a foreign body. The oval dark area overlying the skin in (d) is a marker capsule. [/fig]
|
Role of Cardiologists in the Management of Systemic Lupus Erythematosus: First Reported Case of Three-vessel Disease in a Young Woman in Pakistan
# Introduction
Systemic lupus erythematosus (SLE) is a chronic, autoimmune inflammatory disorder that affects multiple organs of the body, most frequently the kidneys and the heart. Because of the multi-system involvement, SLE has a multitude of presentations including 'butterfly rash', alopecia, arthralgias, painful breathing, seizures and stroke. However, the strongest association of SLE has been found to be with atherosclerosis and coronary artery disease (CAD), with patients of SLE having seven times greater incidence of CAD than those without SLE. Moreover, women of child-bearing age are the populace most affected by SLE and the incidence of major adverse cardiovascular events like myocardial infarction (MI), is much higher in them than in healthy women. Interestingly, the drugs which are used to treat SLE, mainly corticosteroids, are also implicated as risk factors for CAD; thereby, the risk of cardiovascular events in SLE is increased even further. In patients of autoimmune diseases like SLE, cardiovascular events may also be clinically silent in their initial stages. This makes it crucial for these patients to be assessed for a cardiovascular pathology regularly, and the clinical suspicion of such events should always persist even if typical symptoms don't occur.
The antiphospholipid syndrome (APS) is an acquired autoimmune disorder and characterized by hypercoagulability, vascular thrombosis and pregnancy morbidity, with circulating autoantibodies called antiphospholipid (aPL) antibodies, mainly the lupus anticoagulant (LA) and the anticardiolipin (aCL) antibodies. It is referred to as secondary APS when other autoimmune or inflammatory diseases are present. Cardiac manifestations of APS include valvular disease and Libman-Sacks endocarditis primarily, and also occlusive coronary artery disease which can result in angina and myocardial infarction. Here we present a case of a 32-year-old woman with SLE, being treated with steroids, who subsequently developed cardiac manifestations.
## Case presentation
A 32-year-old married woman, known case of SLE, presented to the cardiac emergency department of a tertiary care hospital in Karachi with the complaint of severe chest pain. The patient was diagnosed with SLE in 2006 when she presented with the complaints of chronic fever of unknown origin, joint pain and oral ulcers. Workup at the time revealed positive antidouble-stranded DNA antibody (anti-dsDNA) with a value of over 1000 IU/ml, positive homogeneous anti-nuclear antibodies (ANA), while both anti-smooth muscle antibodies (ASMA) and anti-mitochondrial antibodies (AMA) were negative. C3 and C4 levels in 2012 were 0.61 (normal 0.9-1.8 g/L) and 0.04 (normal 0.1-0.4 g/L) respectively. She was treated with hydroxychloroquine 400 mg/day and prednisolone 10 mg/day in various combinations with several steroid-free periods for the last 13 years. She was diagnosed with hypertension in 2013 which was managed with losartan 50 mg/day.
In May 2019, she presented to us with severe chest pain for 2 hours. The pain was sudden in onset, diffuse and radiating to the left arm. The pain worsened on exertion and started while she was doing her regular activities in the kitchen. She had no associated nausea, vomiting or shortness of breath. She denied any episodes of chest pain in the past. Her physical examination revealed a blood pressure of 140/90 mmHg, a heart rate of 80 beats/minute, a respiratory rate of 16 cycles/min and temperature of 98 °F. An electrocardiogram (ECG) was obtained which showed a rate of 100 beats per minute (BPM), regular sinus rhythm, left axis deviation, and ST-elevation which was present in Lead 1 and aVL with reciprocal changes (STdepression) in Lead II, III and aVF, on the basis of which she was diagnosed with high lateral wall MI.
Clopidogrel 300 mg and aspirin 300 mg were administered as part of the acute coronary syndrome (ACS) protocol following which she underwent an emergency angiography which revealed severe three-vessel disease. The left main coronary artery (LMCA) had a moderate (40-50%) lesion, left anterior descending (LAD) coronary artery had a critical (80-90%) ostial lesion, ramus intermedius branch had a severe (70-80%) lesion, left circumflex coronary artery (LCx) had a moderate (30-40%) lesion and the right coronary artery (RCA) had a severe (80-90%) mid segment lesion. She was kept under observation in the CCU (coronary care unit) and was then transferred to the ICU (intensive care unit) when she was stable. During admission her complete blood count (CBC), urea, creatinine, electrolytes and liver function tests (LFTs) were within normal limits. Erythrocyte sedimentation rate (ESR) and Creactive protein (CRP) were elevated at 66 mm/hour (normal 1-20 mm/hour) and 71.7 mg/L (normal 0.0-2.9 mg/L) respectively. Her total protein was found to be 7 g/dL (normal 6-8 g/dL) of which albumin was 2.9 g/dL (normal 3.5-5 g/dL) and globulin was 4.1 g/dL (normal 2-3.5 g/dL). The albumin to globulin (A/G) ratio was found to be 0.71. Her lipid profile was as follows: total cholesterol level was 177 mg/dL (normal < 200 mg/dL), with HDL (high-density lipoproteins), LDL (low-density lipoproteins), and triglycerides levels being 31 mg/dL (normal 35 mg/dL or more), 113 mg/dL (normal < 100 mg/dL) and 164 mg/dL (normal 50-150 mg/dL) respectively. First troponin I level was 0.219 ng/ml (normal < 1.00 ng/ml) while the second one was 1.89 ng/ml. aCL antibodies were negative, while LA was positive with 1.47 giving a possible diagnosis of APS. Echocardiogram showed normal size left ventricle with normal function, no wall motion abnormalities and no valve abnormalities and a normal left ventricular ejection fraction of 60%.
Post-angiography, she was treated with aspirin 300 mg/day, glyceryl trinitrate 2.6 mg twice a day, metoprolol 50 mg/day, atorvastatin 40 mg/day and hydroxychloroquine 200 mg twice a day. Her maintenance dose of steroids was discontinued as she was found to be in remission.
The cardio-thoracic surgery team was consulted and a coronary artery bypass grafting (CABG) was planned for the patient, which was successfully performed on the 14th of June 2019. Both left internal mammary artery (LIMA) and saphenous vein graft (SVG) were harvested. However, LIMA showed poor flow of blood after grafting, possibly indicating diffuse vascular disease, hence, SVG was grafted to LAD, ramus intermedius branch and RCA successfully.
# Discussion
The involvement of the heart is one of the major causes of morbidity and mortality in people with SLE, although the overall prognosis of the disease itself has improved. Cardiovascular complications of SLE commonly include pericarditis, myocarditis, endocarditis and CAD.
SLE is long known to be one of the strongest risk factors for atherosclerosis and CAD, and studies have shown that the risk of MI is up to nine times greater among SLE patients than the general population, with the risk being 50 times greater in the female patients between the ages of 35 and 44 years.
Increased total cholesterol levels, obesity, aCL antibodies, long duration of corticosteroid use and previous cardiac involvement with SLE are some risk factors that contribute to the development of CAD in patients with SLE. A history of hypertension is also a significant risk factor for the development of CAD in patients with SLE. Although the use of steroids has improved the life expectancy of patients with SLE, this also seems to be causing a rise in the incidence of coronary artery involvement in such patients. This has caused young patients with SLE receiving steroids for their treatment to suffer from angina and MI, which could possibly be due to the formation of atherosclerotic plaques causing stenosis in the coronary vessels that is significant enough to cause clinical symptoms.
Our patient has been a diagnosed case of SLE since 2006 and was on steroid therapy for the past 13 years. She presented to us with typical chest pain and was eventually diagnosed with high lateral wall MI on ECG. Coronary angiography was done which revealed three-vessel disease. On immunological testing she tested negative for aCL antibodies and positive for LA, giving a possible diagnosis of APS. After consulting the cardio-thoracic surgery team, CABG was planned for the patient and was successfully performed. In 2014, a unique case of a 28-year-old young man with an ST-elevation MI was reported. He presented with a history of chest pain and progressive shortness of breath for the past 24 hours. This man was previously healthy with no cardiovascular risk factors. Immunological tests revealed the presence of antibodies for both SLE and APS and the coronary angiography showed total occlusion of LMCA. He was managed with percutaneous coronary intervention (PCI) which was successful, and he was discharged from the hospital. This suggests that the presence of one or more of these aPL antibodies in SLE patients can be an independent risk factor for accelerated atherosclerosis. In literature we find that aPL antibodies are known to be associated with an increased risk of thrombosis and coronary artery occlusion.
Two cases of SLE were reported in 2004, where CABG was performed in women for CAD. The first case was of a 45-year-old woman with SLE for 23 years and receiving steroid treatment for 17 years. She also presented with sub-sternal chest pain, similar to our patient. Coronary angiography revealed 50% narrowing of the LMCA, 70% stenosis in the LAD and 60% stenosis of the circumflex obtuse marginal (OM) 1 artery. CABG was done to LAD and OM 1. The second case was of a 39-year-old woman with SLE, receiving steroids for 17 years whose coronary angiography revealed 80% stenosis in the proximal LAD and CABG was done to LAD. Another case was reported in 2007 of a 42-year-old male hypertensive, diabetic, nonsmoker with SLE. He presented with shortness of breath and class III angina. He was receiving steroid therapy for five years. His coronary angiography was done which revealed a three-vessel disease for which successful CABG was done within two weeks of angiography. These successful cases of CABG surgeries in SLE patients provide corroboration to the fact that a successful CABG has played a potentially life-saving role in our patient as well.
In another case, a 21-year-old female, a diagnosed case of SLE and on steroid therapy for one year presented with chest pain. The cardiac enzymes were normal, and ECG showed STsegment elevations in leads I, aVL, V1-V6 derivations. Her echocardiography revealed hypokinesia in apical-septal, apical-lateral segments of the heart. Interestingly, her angiography showed normal coronary vessels and both aCL antibodies and LA were negative. In comparison to this, in our patient the echocardiography results were normal, however, the angiography revealed three-vessel disease. Furthermore, LA was also positive in our patient.
Another case of a 51-year-old man presenting with dyspnea was reported in 2014. His ECG showed T-wave inversions in the leads II, III, aVF and V4-V6, and the coronary angiography showed total occlusion of LAD and moderate to severe stenosis in the RCA. His blood tests revealed presence of ANA and anti-dsDNA. He was diagnosed as a case of SLE and stable angina with two-vessel disease. He was managed with an urgent off-pump CABG. Although CABG was successful, the patient deteriorated two days after the surgery, and after further vigorous management for active SLE, he was discharged 60 days later. This shows that CABG in patients with SLE is still a challenge due to multiple organs involved and when emergency surgery is required in a patient presenting with both active SLE and CAD, perioperative management is a prerequisite and contributes to a successful surgery. Meticulous peri-operative management was done in our patient before CABG was performed which may have been a contributing factor to the success of her surgery. Another factor could be the difference in the age, our patient being a 32-year-old while this man was a 51-year-old.
# Conclusions
With the advancement in the treatment of SLE, the morbidity and mortality associated with SLE have been decreased, but cardiovascular disease is emerging as an important cause of increased mortality in SLE patients. All patients with SLE, therefore, should be screened for the presence of cardiovascular risk factors and the clinicians should have a low threshold for cardiac evaluation in patients with SLE. This case is the first of its kind to be reported in Pakistan, becoming the prototype for future cases of SLE with cardiac manifestations in the region. The successful outcome of this case can have significant implications regarding the role of cardiologists in the anticipation and management of cardiac disease in SLE. Furthermore, it is safe to imply that the possibility of SLE in a young, especially female patient presenting with symptoms corresponding to CAD, should also be considered by the cardiologists. Therefore, this report can be highly instrumental in highlighting the role of cardiologists in the anticipation and management of cases of SLE with cardiac manifestations.
# Additional information disclosures
Human subjects: Consent was obtained by all participants in this study.
## Conflicts of interest:
In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work. |
Modulation of the functional connectome in major depressive disorder by ketamine therapy
Background. Subanesthetic ketamine infusion therapy can produce fast-acting antidepressant effects in patients with major depression. How single and repeated ketamine treatment modulates the whole-brain functional connectome to affect clinical outcomes remains uncharacterized.Methods. Data-driven whole brain functional connectivity (FC) analysis was used to identify the functional connections modified by ketamine treatment in patients with major depressive disorder (MDD). MDD patients (N = 61, mean age = 38, 19 women) completed baseline resting-state (RS) functional magnetic resonance imaging and depression symptom scales. Of these patients, n = 48 and n = 51, completed the same assessments 24 h after receiving one and four 0.5 mg/kg intravenous ketamine infusions. Healthy controls (HC) (n = 40, 24 women) completed baseline assessments with no intervention. Analysis of RS FC addressed effects of diagnosis, time, and remitter status. Results. Significant differences ( p < 0.05, corrected) in RS FC were observed between HC and MDD at baseline in the somatomotor network and between association and default mode networks. These disruptions in FC in MDD patients trended toward control patterns with ketamine treatment. Furthermore, following serial ketamine infusions, significant decreases in FC were observed between the cerebellum and salience network (SN) ( p < 0.05, corrected). Patient remitters showed increased FC between the cerebellum and the striatum prior to treatment that decreased following treatment, whereas non-remitters showed the opposite pattern. Conclusion. Results support that ketamine treatment leads to neurofunctional plasticity between distinct neural networks that are shown as disrupted in MDD patients. Cortico-striatal-cerebellar loops that encompass the SN could be a potential biomarker for ketamine treatment.
# Introduction
Pharmacotherapies remain the first line of treatment for major depressive disorder (MDD). However, these medications can take several weeks or longer to be effective [bib_ref] Acute and longer-term outcomes in depressed outpatients requiring one or several treatment..., Rush [/bib_ref] and a third of patients, defined as having treatment resistant depression, will remain refractory to two or more treatment trials [bib_ref] What did STAR*D teach us? Results from a large-scale, practical, clinical trial..., Gaynes [/bib_ref] [bib_ref] Prevalence and management of treatment-resistant depression, Nemeroff [/bib_ref]. Over the past decade, ketamine has been identified as a rapidly acting antidepressant [bib_ref] Regulation of neural responses to emotion perception by ketamine in individuals with..., Murrough [/bib_ref] [bib_ref] A randomized trial of an N-methyl-D-aspartate antagonist in treatment-resistant major depression, Zarate [/bib_ref] , with robust therapeutic effects occurring within a day of a single infusion in patients with MDD (aan het [bib_ref] Riluzole for relapse prevention following intravenous ketamine in treatment-resistant depression: A pilot..., Mathew [/bib_ref] [bib_ref] Rapid and longer-term antidepressant effects of repeated ketamine infusions in treatment-resistant major..., Murrough [/bib_ref]. Ketamine is thought to exert its antidepressant effect through N-methyl-D-aspartate (NMDA) receptor antagonism [bib_ref] Antidepressant effects of ketamine in depressed patients, Berman [/bib_ref] [bib_ref] A randomized trial of a low-trapping nonselective N-methyl-D-aspartate channel blocker in major..., Zarate [/bib_ref] α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid modulation, and/or via involvement of opiate receptors or other neurotransmitter systems [bib_ref] The influence of ketamine on drug discovery in depression, Kraus [/bib_ref] [bib_ref] Variations in hippocampal white matter diffusivity differentiate response to electroconvulsive therapy in..., Kubicki [/bib_ref]. Accordingly, the molecular and cellular mechanisms of ketamine's antidepressant action remain under investigation. Furthermore, since most existing studies of the systems-level neurofunctional effects of ketamine [bib_ref] Ketamine normalizes brain activity during emotionally valenced attentional processing in depression, Reed [/bib_ref] have examined single infusion only, whether repeated administration of ketamine engages the same or additional functional brain networks is not known.
MDD is characterized by dysfunctional social, emotional, and reward-based processing, with resultant disruptions in mood, anhedonia (reduced capacity to experience pleasure), and impaired cognitive functioning [bib_ref] A meta-analysis of neurofunctional imaging studies of emotion and cognition in major..., Diener [/bib_ref] [bib_ref] Cognition and depression: Current status and future directions, Gotlib [/bib_ref]. Resting state (RS) networks using functional magnetic resonance imaging (fMRI) have been well studied in MDD [bib_ref] Resting-state functional connectivity in major depression: Abnormally increased contributions from subgenual cingulate..., Greicius [/bib_ref] [bib_ref] Large-scale network dysfunction in major depressive disorder: A meta-analysis of resting-state functional..., Kaiser [/bib_ref]. Prior research has largely focused on the default mode network (DMN), which is most active at rest and is involved in self-referential processing and rumination [bib_ref] Depressive rumination, the default-mode network, and the dark matter of clinical neuroscience, Hamilton [/bib_ref]. However, disturbances in other networks such as salience and cognitive control networks are also reported and proposed to explain depressive symptomology in MDD [bib_ref] Depressive rumination, the default-mode network, and the dark matter of clinical neuroscience, Hamilton [/bib_ref] [bib_ref] Regulation of neural responses to emotion perception by ketamine in individuals with..., Murrough [/bib_ref]. Studies that have investigated RS-fMRI changes 24 h post ketamine infusion in MDD have also shown changes in salience and executive control network, suggesting that ketamine normalizes MDD-related dysfunction in these networks [bib_ref] Ketamine normalizes brain activity during emotionally valenced attentional processing in depression, Reed [/bib_ref]. However, of note, earlier RS-fMRI studies of ketamine have primarily employed seed-based approaches, which only consider a subset of functional brain networks [bib_ref] Reciprocal effects of antidepressant treatment on activity and connectivity of the mood..., Anand [/bib_ref] [bib_ref] Antidepressants normalize the default mode network in patients with dysthymia, Posner [/bib_ref]. Other studies have adopted atlas-based approaches to identify altered networks at the whole brain network level [bib_ref] The connectomics of brain disorders, Fornito [/bib_ref] [bib_ref] Intrinsic connectomes are a predictive biomarker of remission in major depressive disorder, Korgaonkar [/bib_ref] , though this approach is also limited to examining changes in a priori determined atlas-based networks.
Brain networks are generally characterized by a set of brain regions (nodes) and the connections (edges) that link them [bib_ref] Weight-conserving characterization of complex functional brain networks, Rubinov [/bib_ref]. Compared to a priori focused seedbased or atlas-based single-network analysis, whole brain RS-fMRI-based connectomics [bib_ref] Advances and pitfalls in the analysis and interpretation of resting-state FMRI data, Cole [/bib_ref] allow investigation of systems-level treatment effects in depression from the perspective of large-scale brain networks and their complex interactions. Such network modeling can be carried out by parcellating the RS-fMRI into a number of distinct brain regions (nodes) using high-dimensional independent component analysis (ICA), and subsequently estimating the functional connectivity (FC) (i.e. edges) in terms of temporal correlations [bib_ref] Decreased segregation of brain systems across the healthy adult lifespan, Chan [/bib_ref] [bib_ref] Reorganization of brain networks in aging: A review of functional connectivity studies, Sala-Llonch [/bib_ref] [bib_ref] A positive-negative mode of population covariation links brain connectivity, demographics and behavior, Smith [/bib_ref]. The advantage of this method is that it does not require an a priori hypothesis; instead it uses all the information within the data to drive its network estimations. Thus, perturbations in functional networks or the functional connectome relating to antidepressant treatment are rendered more objective and comprehensive. In the current study, we thus used a data-driven approach to investigate the functional connectome of the brain at rest after acute (24 h after first infusion of ketamine) and serial (24 h after receiving four infusions) ketamine treatment in MDD patients. To establish whether treatment-related FC changes normalize toward control values, we also compared FC between controls and MDD patients at baseline. Though employing different methodological approaches, based on previous findings [bib_ref] Ketamine treatment and global brain connectivity in Major depression, Abdallah [/bib_ref] [bib_ref] Default mode connectivity in major depressive disorder measured up to 10 days..., Evans [/bib_ref] , we hypothesized that the network dysfunction in MDD patients would be normalized after ketamine treatment, particularly across the DMN, somato-motor network (SMN), salience network (SN) and fronto-parietal network (FPN). In addition, we aimed to identify correlates of response for MDD patients who reached remission after serial ketamine infusion and hypothesized that the FC changes in the SN would help characterize response.
# Methods and materials
## Subjects
Participants included 40 healthy controls (HC) and 61 DSM-5 defined [SCID] individuals with MDD, who met criteria for treatment resistant depression [i.e. failed two adequate antidepressant trials of adequate dose and duration and had been continuously depressed for ≥6 months in the current episode [bib_ref] Defining treatment-resistant depression, Gaynes [/bib_ref] , all 20-64 years of age]. Subjects were recruited from the Los Angeles area through advertisements, clinician referral, or clinicaltrials.gov (NCT02165449) and received clinical and behavioral assessments and an MRI scan. Patients received a series of four ketamine treatments over a period of 2-3 weeks. Assessments were made at three different time points: (1) initial baseline (TP1) occurring within 1 week of the first treatment (N = 61); (2) 24 h after the first ketamine infusion (TP2, N = 48) and; 24-72 h after the last ketamine infusion (TP3, N = 51). Slightly fewer patients received T2 assessments due to limitations in resources for scanning after the first infusion. At each time point, depression severity was assessed using the Hamilton Depression Rating Scale (HDRS), 17-item [bib_ref] A rating scale for depression, Hamilton [/bib_ref]. Patients whose HDRS score reached ≤7 at the end of treatment (TP3) were considered remitters [bib_ref] Conceptualization and rationale for consensus definitions of terms in major depressive disorder...., Frank [/bib_ref] [bib_ref] A review of studies of the Hamilton depression rating scale in healthy..., Zimmerman [/bib_ref] , whereas the remainder of patients completing four infusions were defined as non-remitters. Demographic and clinical information for all subjects is provided in [fig_ref] Table 1: Patient demographic and clinical information [/fig_ref].
Exclusion criteria for all participants included any unstable medical or neurological condition, current substance abuse, or dependence (ascertained by laboratory testing) or substance abuse history within the preceding 3-months, current or past history of psychosis, schizophrenia, mental retardation, or other developmental disorder, diagnosis of dementia and any contraindication to scanning (e.g. metal implants or claustrophobia). Exclusion criteria for HC also encompassed any current of past psychiatric condition. Prior to treatment, all patients had moderate to severe depressive symptoms as per the HDRS (score ≥17) [bib_ref] A rating scale for depression, Hamilton [/bib_ref]. Patients were also screened to ensure no prior psychotic reactions to medications, alcoholic, or illicit substances in the past, and for other physical or clinical contraindications to ketamine. All subjects provided written informed consent following procedures approved by the University of California, Los Angeles (UCLA) Institutional Review Board (IRB).
## Ketamine treatment
Patients receiving ketamine treatment were permitted to remain on approved monoaminergic antidepressant therapy (if unchanged in the preceding 6-weeks) for the duration of the study. Benzodiazepines were discontinued the night before and morning of all study visits (e.g. scan sessions and ketamine infusion session). Patients received infusions 2-3 times a week for a total of four infusions. At each session, performed as an outpatient procedure, a single sub-anesthetic dose (0.5 mg/kg) of ketamine diluted in 60 cm 3 normal saline was delivered intravenously via pump over a 40-min period in a private room at the UCLA Clinical Research Center or Resnick Neuropsychiatric Hospital. Vital sign monitoring included blood pressure, pulse oximetry, and respiratory rate recording every 3 min and a continuous cardiac rhythm strip. Mental status monitoring also occurred during ketamine infusion to assess for any untoward behavioral or psychological effects.
## Mri data acquisition
A Siemens 3T Prisma MRI system at UCLA's Brain Mapping Center using a 32-channel phased array head coil was used to [bib_ref] Volumetric navigators for prospective motion correction and selective reacquisition in neuroanatomical MRI, Tisdall [/bib_ref]. RS-fMRI data (participants were instructed to focus on a fixation cross) were acquired using two runs of a multiband EPI sequence [bib_ref] Pushing spatial and temporal resolution for functional and diffusion MRI in the..., Ugurbil [/bib_ref] with opposite phase encoding directions [VS = 2 mm isotropic; TR = 800 ms; TE = 37 ms, FA = 52°, MB accl. factor = 8; phase enc. direction = AP (run1)/PA (run2); TA = 6:41 min (per run)].
# Mri data analysis
Anatomical and functional data were visually inspected and minimally preprocessed using the HCP minimal preprocessing pipeline implemented using the BIDS-App [bib_ref] BIDS Apps: Improving ease of use, accessibility, and reproducibility of neuroimaging data..., Gorgolewski [/bib_ref]. Structural and functional artifacts were then removed using a modified sICA + FIX [spatial ICA followed by ICA-based X-noiseifier [bib_ref] Automatic denoising of functional MRI data: Combining independent component analysis and hierarchical..., Salimi-Khorshidi [/bib_ref] ] that utilized concatenation across runs and phase encoding directions of all the fMRI data within a single scanning session . MSMALL alignment was performed to identify and align corresponding cortical locations across subjects based on a combination of cortical folding, thickness, myelination, and RS FC measures [bib_ref] A multi-modal parcellation of human cerebral cortex, Glasser [/bib_ref] [bib_ref] MSM: A new flexible framework for multimodal surface matching, Robinson [/bib_ref] [bib_ref] Multimodal surface matching with higher-order smoothness constraints, Robinson [/bib_ref]. Images with artifacts remaining after pre-processing, which comprised of one subject not counted in the reported sample size, were excluded from the study. Following preprocessing, high-dimensional group-ICA [bib_ref] Probabilistic independent component analysis for functional magnetic resonance imaging, Beckmann [/bib_ref] [bib_ref] Functional segmentation of the brain cortex using high model order group PICA, Kiviniemi [/bib_ref] was performed to generate a set of nodes or parcels. All the RS-fMRI scans were fed into group-level ICA to parcellate the data into a set of 200 spatially-independent components. These spatial maps were projected onto each subject's RS-fMRI time series data to derive one time series per ICA component per subject [dual regression stage 1 [bib_ref] Distinct patterns of brain activity in young carriers of the APOE-epsilon4 allele, Filippini [/bib_ref] ]. After rejecting the group-ICA components based on artifactual noise (i.e. relating either to scanning artifacts, or to non-neuronal biophysical processes such as cardiac fluctuations and head motion), 172 components remained (online. For each subject, the time series associated with these (D = 172) components were used as nodes to perform network analysis [bib_ref] The future of FMRI connectivity, Smith [/bib_ref]. A D × D matrix (netmat) of connectivity estimates was generated using the FSLNets toolbox (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLNets). Network modeling was carried out using partial temporal correlation between node time series. To improve the stability of the connectivity estimates L2 regularization (ρ = 0.1) was applied [bib_ref] Network modelling methods for FMRI, Smith [/bib_ref]. Network matrix or netmat values were converted from Pearson correlation scores (r values) to z statistics using Fisher's transformation. Partial correlation z statistic network matrices were estimated separately for each RS-fMRI run, and then averaged across the runs for each subject, resulting in a single netmat for each subject. The 172 nodes were reordered according to hierarchical clustering of the group-average full correlation netmat using Ward's method implemented in Matlab to generate the functional connectome. Each subject's partial correlation netmat was unwrapped into a single row and combined across subjects to create a Subject × Edges matrix.
# Statistical analysis
The Subject × Edges matrix was used to evaluate group-level analyses including, examination of cross-sectional effects between diagnostic groups at baseline, longitudinal effects of ketamine treatment, and associations between change in neural response Processing pipeline to generate the RS functional connectome. High-dimensional group-ICA and network modeling were performed using FSL-MELODIC and FSLNETS tools. z statistics for the full correlation (below the diagonal) and partial correlation (above the diagonal) were computed for the 172 identified nodes. The nodes were reordered according to a hierarchical clustering of the full correlation matrix. The transformed partial correlations were then arranged in the form of a network matrix (netmat), which was used to perform three main analysis: cross-sectional comparisons of HC and MDD at baseline (TP1), longitudinal comparisons of change between baseline and end of serial treatment in patients (TP1 v. TP3), and comparison of changes between treatment remitters v. non-remitters.
## 2598
Ashish K. Sahib et al.
occurring after single or serial ketamine treatment and antidepressant response. A two-sample t test with age and sex as regressors of no interest compared cross-sectional differences in netmat between HC and MDD at baseline. To test for longitudinal effects of ketamine treatment, paired t tests compared netmats between time points examined pairwise, evaluating effects of both single (TP1-TP2, n = 48) and serial ketamine treatment (TP1-TP3, n = 51). Using a two-sample t test, we also evaluated differences in FC between remitters and non-remitters at baseline (TP1), and the change in FC (ΔFC) after end of treatment (TP1-TP3) with age and sex as regressors of no interest. The FSL randomize tool with 5000 permutations was used for multiple comparisons (controlling family-wise error, FWE) across all edges. Results that were significant at FWE-corrected p < 0.05 are reported. The network edges that survived statistical significance (FWE p < 0.05) were used in a post-hoc analysis using IBM Statistical Packages for the Social Sciences (SPSS v25) to examine relationship between variations in baseline and change in FC with percent change in clinical scores after serial ketamine (TP1-TP3/TP1). Furthermore, FC values were used as dependent measures to confirm the presence of time-by-remission status interactions employing a general linear mixed model with time (TP1 and TP3) and remitter status (remitter, non-remitter) as fixed factors. A p value of <0.05 was used to establish statistical significance in these post-hoc analyses. To test the longitudinal stability of the FC estimates we also compared the FC in a subsample of HC scanned at two time points over a period of 2 weeks (online Supplementary material).
# Results
## Demographic and clinical results
Sex did not differ significantly between HC and MDD groups at baseline, but patients were on average ∼6 years older, p = 0.01 [fig_ref] Table 1: Patient demographic and clinical information [/fig_ref]. To control the variance associated with both sex and age, these variables were included as covariates of no interest in the cross-sectional analysis of FC comparing diagnostic groups.
In MDD patients, HDRS (F (1.89, 91.81) = 83.77, p < 0.0001) scores showed significant improvement across time [fig_ref] Table 1: Patient demographic and clinical information [/fig_ref] , and maximum improvement occurred after serial ketamine infusion (TP3). Of the 51 MDD patients who completed serial ketamine infusions, 24 (47%) achieved remission (HDRS < 7).
## Cross-sectional effects between hc and mdd at baseline
Using p < 0.05 FWE correction and including age and sex as covariates, FC within the somato-motor (nodes 45 and 44) network was found significantly greater in HC as compared to MDD at baseline [fig_ref] Figure 2: Cross-sectional effects [/fig_ref]. Mean FC between these nodes at each time point represented in the bar plots indicates that FC in patients trends toward normalization over time with ketamine treatment. In contrast, FC between the ventral attention node (node 25) and the visual node (node 14) was significantly higher in MDD as compared to controls at TP1. As illustrated in the bar plots in [fig_ref] Figure 2: Cross-sectional effects [/fig_ref] , FC between these nodes significantly reduced with ketamine treatment. Similarly, we also observed higher FC between visual cortex (node 5, encompassing areas V1, V2, and V3) and the DMN node (node 1) in MDD at TP1 that decreased with treatment [fig_ref] Figure 2: Cross-sectional effects [/fig_ref]. There was no significant difference in in-scanner motion metrics at the cross-sectional level as well as over time in MDD subjects (online [fig_ref] Figure 2: Cross-sectional effects [/fig_ref].
## Longitudinal effects of ketamine treatment on the functional connectome
The paired t test comparing the whole-brain FC at baseline (TP1) and after the fourth infusion (TP3) showed significant decrease in FC between a node in the cerebellum (node 49) and a node in the SN (node 32) at p < 0.05 FWE correction . The bar plots show that the average FC between the cerebellum node and the SN node is much higher in MDD patients at baseline (TP1) as compared to controls. Furthermore, this increase in mean FC decreases after the first as well as the fourth infusion of ketamine. No significant differences were found in FC for the paired t test comparing the netmat between TP1 and TP2.
## Fc changes and clinical outcome
At the whole brain level, ΔFC pre-to-post treatment (T1-T3) between remitters and non-remitters showed a significant difference at p < 0.05 FWE correction between a node in the cerebellum (node 126) and a node in the striatum (node 69) [fig_ref] Figure 4: Distinct RS connections representing the difference between change in FC [/fig_ref]. No significant differences were observed between remitters and non-remitters at baseline using p < 0.05 FWE correction. However, there was significant time-by-remission (F (1.969) = 3.742, p = 0.027) status interaction of FC between node 126 and 69. The bar plots in [fig_ref] Figure 4: Distinct RS connections representing the difference between change in FC [/fig_ref] show the average FC reduces over time for remitters, whereas FC for this edge exhibits the opposite trend for non-remitters. When examining average FC for this edge in post-hoc analysis, results also show significantly higher FC for remitters at baseline (TP1) as compared to HC [fig_ref] Figure 4: Distinct RS connections representing the difference between change in FC [/fig_ref].
In addition, there was a significant positive correlation between ΔFC at the end of treatment and %HDRS change (TP1-TP3) after the fourth infusion. Finally, there was a significant positive relationship between FC of the cerebellum node and the SN FC between association and visual network is significantly higher in MDD as compared to controls; node 25: temporal lobe; node 14: visual cortex V1 BA 17. (c) FC between the visual and DMN is significantly higher in MDD as compared to controls; node 5: visual cortex (V1, V2, V3); node 1: right DMN. The color of the diamond connecting the two nodes represents the sign of the group average (patients + HC at TP1) partial correlation (orange: positive, blue: negative). All images are thresholded at z > 8 for visualization. Bar plots show the mean FC for each of the significant ( p < 0.05, FWE-corrected across the netmat) networks for HC and MDD patients at TP1, TP2 and TP3 (*p < 0.05). . Effect of ketamine treatment after fourth infusion (TP1 v. TP3). The FC between the cerebellum and the SN was positively correlated and significantly reduced with ketamine treatment; node 49: cerebellar lobule VI; node 32: secondary somatosensory cortex and insula. The color of the diamond connecting the two nodes represents the sign of the group average (patients + HC at TP1) partial correlation (orange: positive, blue: negative). All images are thresholded at z > 8 for visualization. Bar plots show the mean FC between these nodes for HC and MDD at TP1, TP2 and TP3 (*p < 0.05).
## 2600
Ashish K. node at baseline (TP1) with change in %HDRS (TP1-TP3) after the fourth infusion.
# Discussion
Sub-anesthetic doses of ketamine delivered in either its racemic form of S(+) and R(−) enantiomers [bib_ref] Rapid and longer-term antidepressant effects of repeated ketamine infusions in treatment-resistant major..., Murrough [/bib_ref] [bib_ref] A randomized trial of a low-trapping nonselective N-methyl-D-aspartate channel blocker in major..., Zarate [/bib_ref] , or as (S)-ketamine only [bib_ref] Efficacy and safety of intranasal esketamine adjunctive to oral antidepressant therapy in..., Daly [/bib_ref] [bib_ref] Intravenous esketamine in adult treatment-resistant depression: A double-blind, double-randomization, placebo-controlled study, Singh [/bib_ref] , can reduce depressive symptoms within hours. However, relatively little is known about its effects on brain function at the systems-level following single or repeated ketamine therapy. Furthermore, MDD patients can vary in their response to ketamine treatment with remission rates varying between 29% and 44% [bib_ref] Cognitive behavior therapy may sustain antidepressant effects of intravenous ketamine in treatment-resistant..., Wilkinson [/bib_ref] [bib_ref] A randomized trial of an N-methyl-D-aspartate antagonist in treatment-resistant major depression, Zarate [/bib_ref]. Whether patterns of FC can predict or elucidate the neurofunctional mechanisms associated with remission following ketamine therapy remains mostly unaddressed. Moreover, previous FC studies in MDD have to date exclusively focused on investigating specific brain networks such that other changes in the functional connectome associated with ketamine may be missed. The current study thus leveraged a data-driven network modeling approach to investigate how single and repeated ketamine administration modulates this functional connectome in MDD. To this end, 172 distinct functional nodes were generated from a highdimensionality group-ICA that map several large-scale networks including auditory, somatosensory, motor, and visual activity as well as higher cognitive processes like executive function [bib_ref] Distinct resting-state functional connections associated with episodic and visuospatial memory in older..., Suri [/bib_ref]. Using these nodes/networks, distinct RS connections were mapped to generate a functional connectome. Changes in the functional connectome across time in patients, and between controls and MDD patients prior to ketamine infusion were subsequently addressed.
At the cross-sectional level, results revealed that FC within the SMN was positively correlated for controls as compared to MDD patients, where it was anticorrelated. This anticorrelation in the SMN decreased in MDD patients with ketamine treatment. Furthermore, we also showed that the FC between the VN and FPN/association network was positively correlated in MDD patients, whereas it was anticorrelated in HC. These positive correlations between the FPN/association network and VN decreased in MDD patients with ketamine treatment. These findings suggest that the disrupted network connections that we observe in MDD patients at baseline are reaching normative (control) values with ketamine treatment. Specifically, the distinct RS connection that involved the cerebellum and the SN was uniquely coupled with repeated ketamine administration. The FC between these nodes was positively correlated at baseline in MDD and much greater than HC. With repeated ketamine infusion the FC in this network decreased, again trending toward normative values. In terms of clinical response remitters showed increased positive correlation at baseline between the cerebellum and the striatum that significantly decreased with ketamine treatment. But the non-remitters showed an opposite trend that involved the FC to be similar that of controls and increased in strength at the end of ketamine treatment. Overall, these findings suggest that repeated low-dose ketamine treatment normalizes the dysfunction of the SN in MDD. In particular, the cortico-striatal-cerebellar (CSC) loops that encompass the SN could be a potential biomarker for ketamine treatment.
## Cross-sectional effects: hc v. mdd
Patients with MDD are known to show reduced FC within and across several brain networks at rest. Consistent with previous findings, MDD patients at baseline showed lower mean FC in the SMN as compared to controls. This reduced FC within the SMN may be linked with psychomotor retardation, which has been characterized as a key feature of patients with major depression [bib_ref] Psychomotor retardation in depression: Biological underpinnings, measurement, and treatment, Buyukdura [/bib_ref] [bib_ref] Localized connectivity in depression: A meta-analysis of resting state functional imaging studies, Iwabuchi [/bib_ref]. A recent study also showed decreased FC in the SMN prior to treatment in MDD, which increased with standard pharmacotherapeutic treatment [bib_ref] Intrinsic connectomes are a predictive biomarker of remission in major depressive disorder, Korgaonkar [/bib_ref]. Although not significant, in the current study we observed a similar effect of FC within this SMN in MDD patients with ketamine treatment. These results suggest that ketamine tends to normalize the dysfunction in this SMN that may be associated with its antidepressant effect. Furthermore, the FC between the VN and the DMN/association network was anticorrelated in controls, whereas MDD patients showed an opposite effect. Prior studies have reported disrupted FC within the VN and the DMN/association networks [bib_ref] Large-scale network dysfunction in major depressive disorder: A meta-analysis of resting-state functional..., Kaiser [/bib_ref] [bib_ref] Reduced default mode network functional connectivity in patients with recurrent major depressive..., Yan [/bib_ref]. Here, for the first time we present disrupted network connectivity across these networks as well. In addition, ketamine is known to modulate metabolism in the occipital cortex [bib_ref] Neural correlates of rapid antidepressant response to ketamine in treatment-resistant unipolar depression:..., Carlson [/bib_ref] , and visual cortex activity has been shown to correlate with antidepressant response [bib_ref] Potential of pretreatment neural activity in the visual cortex during emotional processing..., Furey [/bib_ref]. Notably, we have also previously found cerebral blood perfusion changes in primary and secondary visual areas following ketamine therapy to be associated with antidepressant response in an overlapping sample of patients . Similarly, at the end of ketamine treatment we observed a significant modulation of FC between VN and the association network [fig_ref] Figure 2: Cross-sectional effects [/fig_ref] with treatment. These changes showed a trend toward controls, indicating the antidepressant effects of ketamine include restoration of FC.
## Ketamine modulation of the functional connectome in mdd
Relative to baseline, 24 h after the fourth infusion of ketamine we observed a significant decrease in FC between the node in the dorsal salience (node 32) networkand the node in the SMN part of the cerebellum (node 49, [bib_ref] Distinct cerebellar contributions to intrinsic connectivity networks, Habas [/bib_ref]. The cerebellum participates in motor as well as nonmotor functions [bib_ref] Distinct cerebellar contributions to intrinsic connectivity networks, Habas [/bib_ref] and MDD pathophysiology affects distinct sub-regions of the cerebellum that communicate with cortical networks, thereby sub-serving several cognitive and mood functions [bib_ref] Cerebellar contributions to Major depression, Depping [/bib_ref]. Furthermore, cerebellar NMDA receptors play a vital role in motor coordination and motor learning [bib_ref] Modulation of NMDA receptors in the cerebellum. II. Signaling pathways and physiological..., Sanchez-Perez [/bib_ref] and prior studies have shown that NMDA receptor antagonists restore motor activity in mutant nice [bib_ref] Impairment of N-methyl-D-aspartate receptorcontrolled motor activity in LYN-deficient mice, Umemori [/bib_ref].
Patients with MDD are known to have abnormal functioning of the SN [bib_ref] Large-scale network dysfunction in major depressive disorder: A meta-analysis of resting-state functional..., Kaiser [/bib_ref] [bib_ref] Resting-state functional MRI in depression unmasks increased connectivity between networks via the..., Sheline [/bib_ref]. The SN plays a critical role in switching between task negative network (DMN) and the task positive (FPN) [bib_ref] Saliency, switching, attention and control: A network model of insula function, Menon [/bib_ref]. As a result, any abnormalities in the SN can affect several cognitive functions [bib_ref] An information theoretical approach to prefrontal executive function, Koechlin [/bib_ref] [bib_ref] An integrative theory of prefrontal cortex function, Miller [/bib_ref]. In particular, the region reported in this study that encompasses the SN is implicated in patients with posttraumatic stress disorder. Notably, the dysfunctions associated with the SN in MDD are also found to be normalized with ketamine treatment in prior studies [bib_ref] Ketamine treatment and global brain connectivity in Major depression, Abdallah [/bib_ref] [bib_ref] Default mode connectivity in major depressive disorder measured up to 10 days..., Evans [/bib_ref]. Although not significant, the mean FC between the node in the cerebellum and the SN in patients with MDD was greater as compared to HC. In accordance with previous ketamine studies [bib_ref] Ketamine treatment and global brain connectivity in Major depression, Abdallah [/bib_ref] [bib_ref] Default mode connectivity in major depressive disorder measured up to 10 days..., Evans [/bib_ref] , we show that there are disruptions within as well as across the SMN and SN in patients with MDD. This disruption across the SMN and the SN normalizes toward controls with repeated ketamine therapy further suggesting that neuroplasticity across these networks contributes to the antidepressant effects of ketamine at the brain systems level. A recent study that employed a data-driven approach of global brain connectivity (GBC) has also shown that early GBC in the SMN predicts response to sertraline treatment. Furthermore, a recent study from our group has also shown that functional plasticity in the SMN after ketamine treatment relates to improvements in depressive symptoms, suggesting modulation of this network plays an important role in therapeutic response.
## Fc associated with treatment response
Almost half of the patients included in the current study remitted at the end of treatment. Our MDD cohort overall had a significant higher FC of the SN at baseline . However, on splitting the group based on clinical outcome, only remitters showed a significantly higher FC between the cerebellum node and a node in the striatum. The basal ganglia are essential for reward-based learning (Watabe-Uchida, Eshel, & Uchida, 2017) and imaging studies have also indicated brain regions outside the basal ganglia, including the cerebellum, to be associated with reward processing [bib_ref] Prediction error in reinforcement learning: A meta-analysis of neuroimaging studies, Garrison [/bib_ref] [bib_ref] Human neural learning depends on reward prediction errors in the blocking paradigm, Tobler [/bib_ref]. Furthermore, lesions in the cerebellum are associated with impairments in reinforcement-learning [bib_ref] The cerebellum is involved in reward-based reversal learning, Thoma [/bib_ref]. It should also be noted that the cerebellum and the basal ganglia are interconnected at the sub-cortical level [bib_ref] The basal ganglia communicate with the cerebellum, Bostan [/bib_ref] [bib_ref] An HRP study of hypothalamo-cerebellar and cerebello-hypothalamic connections in squirrel monkey (Saimiri..., Haines [/bib_ref] [bib_ref] The cerebellar-hypothalamic axis: Basic circuits and clinical observations, Haines [/bib_ref] and increased activity is observed in this circuit across various neurological disorders [bib_ref] The basal ganglia and the cerebellum: Nodes in an integrated network, Bostan [/bib_ref]. This CSC loop along with the insula has been shown to be a key neural circuit of the SN [bib_ref] Distinct cerebellar contributions to intrinsic connectivity networks, Habas [/bib_ref]. In line with these findings, compared to controls, remitters in the current study showed significantly higher FC in the CSC loop as compared to controls at baseline, which significantly decreased with ketamine treatment. In contrast, non-remitters showed lower FC as compared to controls that increased with ketamine treatment. In addition, we observed a significant relationship between change in FC in this CSC circuit with improvement in clinical outcome, and baseline FC in this loop could predict end of treatment outcome. Taken together and in line with previous findings, these observations suggest that SN dysfunction in patients with MDD, particularly those more prone to reward-based deficits (CSC loops) might respond better to ketamine treatment and further studies of basal ganglia-cerebellar-cerebral cortical networks could help better understand the pathophysiology of this disorder.
# Limitations
Several limitations should be acknowledged for the current investigation. First, the current study was not a randomized clinical trial and did not include an active control condition. A previous study has implicated RS in predicting placebo effects [bib_ref] Salience network functional connectivity predicts placebo effects in Major depression, Sikora [/bib_ref]. However, it is important to note that the focus of this mechanistic clinical trial was to investigate the perturbation of the functional connectome at rest associated with ketamine rather than to address clinical efficacy, and patients serve as their own
## 2602
Ashish K. Sahib et al.
controls when examining change in FC over time. Furthermore, changes in FC over the course of ketamine treatment related to antidepressant response suggest that these changes are not reflective of placebo effects. Also, a subsample of HC subjects was scanned twice to estimate normative variance in FC over time (online Supplementary material). Notably, none of the edges that were significantly modulated with ketamine treatment showed a change in FC for HC over time in the direction of ketamine treatment (online . These results are in line with other studies that demonstrate the patterns of RS brain activity measures are relatively stable across individual control subjects scanned multiple times [bib_ref] Variability and reliability of effective connectivity within the core default mode network:..., Almgren [/bib_ref] [bib_ref] Group independent component analysis reveals consistent resting-state networks across multiple sessions, Chen [/bib_ref]. Though cross-sectional changes between MDD and HC subjects were investigated to assist with the interpretation of ketamine findings, it is possible that the inclusion of patients with comorbid psychiatric diagnoses as well as the heterogeneity of symptoms among depressed individuals may have influenced the observed results between diagnostic groups. Finally, patient participants were allowed to continue concurrent stable anti-depressant medication, which may have impacted findings.
# Conclusion
The use of data-driven functional connectomics to identify brain networks at rest is a major methodological advancement that could translate to improved strategies for diagnosis and for tailoring treatment in patients with MDD. Using this approach our findings provide novel insights into the specific components of large-scale brain networks associated with ketamine therapy. We have identified disrupted FC within the SMN, and between the visual and association networks in MDD patients that normalize with ketamine treatment. Furthermore, we found distinct changes in RS FC in SN networks with respect to ketamine treatment. Finally, in terms of remission, we found greater than normal FC in the CSC loop in patients who remit, which may be a prerequisite mechanism for ketamine treatment. Future studies may expand upon the current findings, including addressing how treatment-related changes in the functional connectome vary in relation to longer term clinical outcomes, and maintenance of therapeutic response.
Supplementary material. The supplementary material for this article can be found at https://doi.org/10.1017/S0033291720004560
[fig] Figure 1: (a) Study design illustrating the timing of MRI sessions and clinical assessments relative to ketamine infusions. (b) [/fig]
[fig] Figure 2: Cross-sectional effects: controls v. MDD. (a) FC between nodes in the SMN is significantly lower in MDD as compared to controls; node 45: premotor cortex, Brodmann area (BA) 6; node 44: primary motor cortex, BA 4a. (b) [/fig]
[fig] Figure 4: Distinct RS connections representing the difference between change in FC (TP1-TP3) for remitters and non-remitters. (a) The FC between the cerebellum node and the node in the striatum along with insula showed the largest change in FC between remitters and non-remitters: node 69: cerebellar left crus 1; node 126: left putamen and insula. The color of the diamond connecting the two nodes represents the sign of the group average (patients + HC at TP1) partial correlation (orange: positive, blue: negative). All images are thresholded at z > 8 for visualization. (b) Bar plots show the mean FC between these nodes for HC, remitters and non-remitters at TP1, TP2 and TP3 (*p < 0.05). (c) ΔFC (TP1-TP3) between the cerebellum and striatum showed a significant positive relationship with %HDRS change (TP1-TP3). (d ) Baseline (TP1) FC between the cerebellum and the striatum showed a significant positive relationship with %HDRS change (TP1-TP3). [/fig]
[table] Table 1: Patient demographic and clinical information [/table]
|
Legacy effects of short-term intentional weight loss on total body and thigh composition in overweight and obese older adults
OBJECTIVE: Weight regain following intentional weight loss may negatively impact body composition, accelerating fat regain and increasing risk of physical disability. The purpose of this study was to compare long-term changes in whole body and thigh composition in obese older adults who intentionally lost and then partially regained weight to obese older adults who remained weight stable. SUBJECTS/METHODS: This pilot study analyzed total body (dual-energy X-ray absorptiometry (DXA)) and thigh (computed tomography (CT)) composition data collected from 24 older (65-79 years) adults 18 months after completion of a 5-month randomized trial that compared resistance training alone (RT) with RT plus caloric restriction (RT+CR). RESULTS: Mean loss of body mass in the RT+CR group (n = 13) was 7.1 ± 2.4 kg during the 5-month intervention (74% fat mass; 26% lean mass; all P o 0.01), whereas RT (n = 11) remained weight stable (+0.3 ± 1.8 kg; P = 0.64). Differential group effects were observed for all DXA and CT body composition measures at 5 months (all P ⩽ 0.01); however, by 23 months, group differences persisted only for total body (RT+CR: 81.6 ± 10.0 kg vs RT: 88.5 ± 14.9 kg; P = 0.03) and lean (RT+CR: 50.8 ± 9.3 kg vs RT: 54.4 ± 12.0 kg; P o 0.01) mass. All RT+CR participants regained weight from 5 to 23 months (mean gain = +4.8 ± 2.6 kg; P o 0.01). Total fat mass and all thigh fat volumes increased, whereas thigh muscle volume decreased, during the postintervention follow-up in RT+CR (all P ⩽ 0.01). In the RT group, body mass did not change from 5 to 23 months (−0.2 ± 0.9 kg; P = 0.87). Decreased total thigh volume, driven by the loss of thigh muscle volume, were the only postintervention body composition changes observed in the RT group (both P o0.
# Introduction
Obesity and aging are major risk factors for the development and recurrence of a wide array of chronic disease. [bib_ref] Obesity in older adults: technical review and position statement of the American..., Villareal [/bib_ref] [bib_ref] The public health impact of obesity, Visscher [/bib_ref] In addition, aging is associated with increased adiposity, including preferential fat deposition in ectopic regions, [bib_ref] Adipose tissue, diet and aging, Zamboni [/bib_ref] such as within the skeletal muscle. [bib_ref] Longitudinal study of muscle strength, quality, and adipose tissue infiltration, Delmonico [/bib_ref] The amount of fat stored in and around thigh muscle, especially intermuscular fat, is associated with reduced muscle function and strength. [bib_ref] Leg muscle mass and composition in relation to lower extremity performance in..., Visser [/bib_ref] [bib_ref] Muscle mass, muscle strength, and muscle fat infiltration as predictors of incident..., Visser [/bib_ref] [bib_ref] Associations between body composition and gait-speed decline: results from the Health, Aging,..., Beavers [/bib_ref] When coupled with the well-documented age-related loss of muscle mass (e.g., sarcopenic obesity), these changes in body composition may accelerate declines in physical function and increase risk of physical disability. [bib_ref] Obesity, sarcopenia and their functional consequences in old age, Visser [/bib_ref] Lifestyle-based weight loss interventions in obese older adults, including caloric restriction and exercise, can be very effective in reducing total body and ectopic fat stores, at least in the short term. However, weight regain following intentional weight loss is common, [bib_ref] Maintenance of weight loss after lifestyle interventions for overweight and obesity, a..., Barte [/bib_ref] and emerging evidence suggests that it may negatively influence body composition in older adults and perhaps promote sarcopenic obesity. As we [bib_ref] Is lost lean mass from intentional weight loss recovered during weight regain..., Beavers [/bib_ref] [bib_ref] Cardiometabolic risk after weight loss and subsequent weight regain in overweight and..., Beavers [/bib_ref] and others [bib_ref] A losing battle: weight regain does not restore weight loss-induced bone loss..., Villalon [/bib_ref] [bib_ref] Early weight regain after gastric bypass does not affect insulin sensitivity but..., Tamboli [/bib_ref] have shown, older adults show a faster accumulation of total fat vs muscle mass during weight regain after intentional weight loss. Observational data of older adults also show that proportionally more muscle mass is lost during periods of weight loss than is gained during periods of weight gain or regain. [bib_ref] Composition (lean and fat tissue) of weight changes in adult Danes, Heitmann [/bib_ref] [bib_ref] Weight change and the conservation of lean mass in old age: the..., Newman [/bib_ref] [bib_ref] Weight loss and regain and effects on body composition: the Health, Aging,..., Lee [/bib_ref] However, these prior studies do not discriminate body composition changes because of a cycle of weight loss and regain from longitudinal aging-related changes. The 'lack of adjustment for the age-related decline in fat-free mass' is cited as a major limitation to full understanding of the ramifications of weight cycling on body composition in older adults. [bib_ref] Measuring the impact of weight cycling on body composition: a methodological challenge, Bosy-Westphal [/bib_ref] As this population is also more likely to store excess energy as fat in ectopic regions, [bib_ref] Longitudinal study of muscle strength, quality, and adipose tissue infiltration, Delmonico [/bib_ref] [bib_ref] Pericardial fat, visceral abdominal fat, cardiovascular disease risk factors, and vascular calcification..., Rosito [/bib_ref] [bib_ref] Aging and regional differences in fat cell progenitors-a mini-review, Sepe [/bib_ref] [bib_ref] Age Related Shift in Visceral Fat, Hunter [/bib_ref] it is also important to determine whether weight regain following weight loss results in increased relative fat storage in these less desirable locations.
Thus, the purpose of this study was to compare whole body and thigh composition in older adults who intentionally lost and then partially regained weight to older adults who remained weight stable. We accomplished this by recalling participants 18 months after completion of a 5-month randomized trial designed to compare a resistance training and caloric restriction (RT+CR) intervention to resistance training alone (RT). We also compared changes in whole body and thigh composition within groups, highlighting changes that occurred with weight loss and subsequent weight regain.
# Materials and methods
## Study overview
This pilot study collected and analyzed whole-body and thigh composition data 18 months after completion of a 5-month randomized controlled trial, originally designed to compare the effects of RT+CR to RT alone on change in body composition and skeletal muscle function. The parent study design, recruitment and main outcome data are published. [bib_ref] Effects of resistance training with and without caloric restriction on physical function..., Nicklas [/bib_ref] Briefly, men and women were recruited and enrolled in the parent study on the basis of the following criteria: (a) age 65-79 years, (b) no resistance training in the past 6 months, (c) body mass index = 27-34.9 kg m −2 , (d) non-smoking for the past year, (e) weight stable ( o5% weight change), (f) normal cognitive function and, (g) no evidence of clinical depression, heart disease, cancer, liver or renal disease, chronic pulmonary disease, uncontrolled hypertension, physical impairment or other contraindication for exercise.
After completion of the 5-month trial, a subsample of participants in the RT+CR group who lost ⩾ 5% of their body weight during the intervention (n = 13) and participants in the RT group who lost o5% of their body weight (n = 11) and who had complete DXA and CT measurements were contacted sequentially until we reached our a priori enrollment sample size (n = 24). This study was approved by the Wake Forest School of Medicine Institutional Review Board, all experiments were conducted in accordance with the Declaration of Helsinki and participants provided written informed consent.
## Resistance training and caloric restriction intervention descriptions
Details of the RT and CR interventions are published. [bib_ref] Effects of resistance training with and without caloric restriction on physical function..., Nicklas [/bib_ref] Briefly, all participants in the study underwent 5 months of supervised RT 3 days per week on resistance machines. The machines used were: (1) leg press; (2) leg extension; (3) seated leg curl; (4) seated calf; (5) incline press; (6) compound row; (7) triceps press; and (8) bicep curl. The maximal weight that a person could lift with correct form in a single repetition (1RM) was used to prescribe intensity and the training goal was to complete three sets of 10 repetitions for each exercise at 70% 1RM for that specific exercise. Resistance was increased when a participant was able to complete 10 repetitions on the third set for two consecutive sessions. Strength testing was repeated every 4 weeks and training loads were adjusted to be consistent with the 70% 1RM goal. Each participant recorded the weight lifted, number of repetitions completed and number of sets completed for each exercise.
Participants in the RT+CR group performed the same RT program and also underwent CR for 5 months. The CR protocol used meal replacements (2 per day), nutrition education sessions (weekly) and dietary modification advice (as needed), targeting a 600 kcal per day reduction from estimated daily energy needs for weight maintenance. To verify compliance with the CR intervention protocol, participants were asked to keep a diet log of all foods consumed and the logs were monitored weekly by the study registered dietitian.
## Total body and thigh composition measures
Body composition measures were obtained at baseline, 5, and 23 months. Height and body mass were measured without shoes and outer garments removed. Body mass index was calculated as body mass (kg) divided by height squared (m 2 ). Whole-body fat and lean mass (kg) were measured by dual-energy X-ray absorptiometry (DXA; Hologic Delphi QDR, Bedford, MA, USA) following standard procedures. Computed tomography (CT) scans, using a GE 16-slice Light Speed Pro (GE Medical Systems, Milwaukee, WI, USA), quantified total thigh volume, thigh muscle volume and thigh fat (total, subcutaneous and intermuscular) fat volume (cm 3 ). Participants were placed supine in the scanner with arms above the head and legs flat. Scans were conducted at 120 KVp, 350 mA, 10-mm helical with a pitch of 11.25 mm per rotation and a gantry speed of 0.8 s. Thigh muscle and adipose tissue volumes were measured using a 5-cm section of the thigh centered at the junction of the proximal and middle-third of the femur as measured from the scout topogram. The volume of muscle and adipose tissue was segmented and measured using the GE Healthcare, Advantage Windows 4.2 Volume Viewer (GE Healthcare, Waukesha, WI, USA). Thigh muscle area was considered the total area of non-adipose and non-bone tissue within the deep fascial plane. For adipose tissue volume, sequences were reconstructed into the maximum 50 cm field-of-view to prevent truncation of subcutaneous fat on larger individuals and a 2.5-mm, no-gap, slice thickness was used. Intermuscular adipose tissue was separated from subcutaneous adipose tissue by drawing a line along the deep fascial plane surrounding the thigh muscles.
# Statistical analysis
Continuous body composition measures were summarized at each time point using means and standard deviations. Randomization group means at the 5-and 23-month assessment time points were modeled using mixed linear models fit with randomization, time and their interaction adjusted for baseline values of the outcome. Visit-specific group comparisons were performed using contrast statements at each follow-up visit. Absolute and percent changes in body weight, body composition and fat distribution variables were calculated as the baseline value subtracted from the 5-month value and 23-month value subtracted from 5-month value. Changes in body composition between baseline and 5 months and between 5 and 23 months were modeled using mixed linear models adjusted for baseline value of the outcome. Analyses were performed using the SAS software (version 9.4; SAS Institute, Cary, NC, USA) and P ⩽ 0.05 was considered statistically significant.
# Results
The mean baseline age of all participants was 70 ± 4 years, 50% were female (n = 12) and the majority were non-Hispanic white (n = 23). All participants were overweight or obese at baseline, with an average BMI of 29.9 ± 2.0 kg m −2 (RT+CR = 29.2 ± 1.5 vs RT = 30.7 ± 2.3 kg m −2 ).
Between-and within-group differences in total body mass and composition Means (± s.d.) for total body mass and composition at baseline, at the end of the 5-month intervention and at the 23-month followup visit are presented by group in . RT and RT+CR groups were balanced at baseline with respect to total body mass and composition. RT+CR participants lost 7.1 ± 2.4 kg of body mass (6.3 ± 3.4%; P o 0.01) during the intervention, with~74% of mass lost as fat mass and 26% as lean mass (both P o 0.01). By design, total body mass and composition did not change significantly in the RT group during the 5-month study, and both total body fat and lean mass were significantly higher in the RT group compared with the RT+CR at the 5-month time point (both P o 0.01). By 23 months, body mass and lean mass were still lower (P o0.01) in the RT+CR vs the RT group, but fat mass was not significantly different between groups.
Between-and within-group differences in thigh composition Thigh composition variables by group and time point are presented in [fig_ref] Table 2: Thigh composition variables by group and time point [/fig_ref] , with groups balanced across all variables at baseline, save total thigh volume (1548.6 ± 144.4 vs 1376.9 ± 122.2 cm 3 in the RT and RT+CR groups, respectively). Upon completion of the 5-month trial, all thigh compartment volumes were lower in the RT+CR group compared with the RT group. Total thigh fat volume (driven by decreases in all fat compartments) decreased in the RT+CR group (all P ⩽ 0.01), whereas thigh muscle volume did not change during intervention in this group. In the RT group, thigh muscle volume increased bỹ 5% with the intervention, whereas thigh fat volumes were unchanged from baseline. Between-group 23-month comparisons showed no lasting intervention-related differences for any thigh composition measurements.
Within-group changes in total body and thigh composition during postintervention follow-up All RT+CR participants regained some body mass from 5 to 23 months (average: +4.8 ± 2.6 kg; range: 0.4-9.1 kg); however, total body mass was still lower than baseline (83.9 ± 11.5 vs 81.6 ± 10.0 kg; P o0.01). In the RT group, body mass did not change during follow-up (−0.2 ± 0.9 kg; P = 0.87). Given the interest in studying the effect of weight regain following intentional weight loss on change in total body and thigh composition,and lean compartments during and after the intervention for the RT+CR group. As shown in, less absolute fat mass was regained during follow-up (+4.4 ± 0.6 kg from 5 to 23 months) than was lost during intervention (−5.1 ± 0.3 kg from baseline to 5 months), whereas lean mass loss was observed both during and after the intervention (−1.8 ± 0.4 kg from baseline to 5 months and − 0.3 ± 0.4 kg from 5 to 23 months) in RT+CR. Thus, a slightly greater percentage of the total mass that was regained was comprised of fat compared with the percentage of total mass that was lost as fat. There were no significant changes in fat or lean mass during months 5 to 23 in the RT-only group (see [fig_ref] Table 2: Thigh composition variables by group and time point [/fig_ref]. During the 18-month follow-up, RT+CR participants continued losing thigh muscle volume (−16.1 ± 5.6 cm 3 ) and regained thigh fat volume (+83.5 ± 18.7 cm 3 ;. Examination of changes in thigh intermuscular and subcutaneous fat volumes shows that most of the fat loss was from the subcutaneous compartment, and that~82% of lost thigh subcutaneous fat was regained during the follow-up period (−94.7 ± 16.7 cm 3 from baseline to 5 months vs +78.1 ± 18.5 cm 3 from 5 to 23 months). Although a smaller amount of thigh intermuscular fat volume was lost with weight loss, nearly 100% was regained with weight regain (−5.5 ± 1.6 cm 3 from baseline to five months vs +5.4 ± 1.4 cm 3 from 5 to 23 months;. Thigh muscle volume, and thus total thigh volume, decreased during the 18-month follow-up in the RT-only group (both P o 0.04), but there were no significant changes in any of the thigh fat volumes from months 5 to 23 (see [fig_ref] Table 2: Thigh composition variables by group and time point [/fig_ref].
# Discussion
The primary purpose of this study was to explore the legacy effects of a short-term RT+CR intervention, compared with RT alone that did not result in weight loss, on total body and thigh composition in overweight and obese, older adults. Our data indicate that despite clinically meaningful weight loss and concurrent favorable shifts in total body and thigh composition in participants who lost 45% of initial body mass by undergoing CR during RT, these improvements were generally not sustained long term. Indeed, at the 23-month follow-up visit, in spite of modestly reduced total body mass in the RT+CR group, total fat mass and thigh fat volumes (including intermuscular fat) were not different from participants who did not lose weight during RT alone. Furthermore, changes in body composition with weight regain during the follow-up show that there is total and regional fat accretion, but loss of lean tissue (particularly thigh muscle volume), implying greater long-term risk for sarcopenic obesity in those who lost weight vs those who did not during RT.
Observational cohort data in older adults examining body composition changes with weight loss and subsequent regain suggest that weight cycling may accelerate loss of lean mass with aging. [bib_ref] Weight change and the conservation of lean mass in old age: the..., Newman [/bib_ref] [bib_ref] Weight loss and regain and effects on body composition: the Health, Aging,..., Lee [/bib_ref] This finding is echoed in the few long-term follow-up studies of RCTs that resulted in weight loss, where weight regain following significant intentional weight loss is shown to be comprised predominantly of fat mass. [bib_ref] Is lost lean mass from intentional weight loss recovered during weight regain..., Beavers [/bib_ref] [bib_ref] Longterm maintenance of weight loss after lifestyle intervention in frail, obese older..., Waters [/bib_ref] [bib_ref] This work is licensed under a Creative Commons Attribution 4.0 International License...., Pownall [/bib_ref] For example, we previously showed that, in postmenopausal women, for every 1 kg of fat mass lost during weight loss, 0.26 kg of lean tissue was lost, whereas for every 1 kg of fat mass regained in the following year, only 0.12 kg of lean tissue was regained. [bib_ref] Is lost lean mass from intentional weight loss recovered during weight regain..., Beavers [/bib_ref] Similarly, 30 months after completion of a 1-year weight loss intervention in frail, obese older adults, despite continued reduced total body mass from baseline, about one-half (45%) of total fat mass had been regained but lost lean mass was not recovered. [bib_ref] Longterm maintenance of weight loss after lifestyle intervention in frail, obese older..., Waters [/bib_ref] Last, recent long-term data from the Look AHEAD trial examining changes in body composition over an 8-year follow-up period demonstrate that although a lifestyle-based caloric restriction and exercise program produced significant fat and lean mass loss during the first year of intervention, 7 years later, 100% of fat mass was regained, whereas participants continued to lose lean mass. [bib_ref] This work is licensed under a Creative Commons Attribution 4.0 International License...., Pownall [/bib_ref] However, none of these prior studies included a weight-stable control group in order to determine whether weight cycling is causal for these adverse changes in body composition. Our data show that, while both groups lost lean tissue during the postintervention follow-up period, the group that lost and regained weight (e.g., RT+CR) also experienced a gain in fat during this same period. Thus, these pilot data suggest that fat, rather than muscle, tissue is regained when older adults regain weight after weight loss. Over time, and/or with repeated bouts of weight cycling, this could exacerbate sarcopenic obesity in older adults. Importantly, our results build on these existing published studies because, to the best of our knowledge, this is the first assessment of change in thigh composition with weight regain in older adults. These data may be even more clinically relevant as fat deposition in and around the thigh muscle (especially intermuscular fat) is an important body composition predictor of decline in gait speed, loss of muscle strength and onset of mobility disability among older adults. [bib_ref] Leg muscle mass and composition in relation to lower extremity performance in..., Visser [/bib_ref] [bib_ref] Muscle mass, muscle strength, and muscle fat infiltration as predictors of incident..., Visser [/bib_ref] [bib_ref] Associations between body composition and gait-speed decline: results from the Health, Aging,..., Beavers [/bib_ref] Group differences presented in this pilot study show that weight loss via CR during RT results in lower total, subcutaneous and intermuscular thigh fat, as well as lower thigh muscle volumes, compared with RT without weight loss immediately after intervention cessation. However, by 18 months postintervention, despite those who underwent caloric restriction maintaining a lower body mass, none of the group differences in thigh composition were evident; there was a trend (P = 0.11) for muscle volume to still be higher in RT vs RT+CR. Notably, during the postintervention follow-up, RT+CR lost thigh muscle volume and partially regained total (81%) and subcutaneous (82%) thigh fat, but regained 100% of intermuscular fat. Thus, even though (on average) RT+CR participants did not regain all lost weight, fat mass or total thigh fat volume by 18 months, thigh muscle volume was lower and all of lost thigh intermuscular fat was regained.
Strengths of the current study consist of: (1) inclusion of a weight-stable control group, (2) long-term follow-up measures of total body and thigh composition using imaging techniques and (3) follow-up assessment of the subset of participants who lost a clinically significant amount of weight. Weaknesses include the potential selection bias which may have occurred because of our sampling methods. Because of the limited sample size and resources of this pilot project, a decision was made a priori, to ensure differential weight loss was achieved in RT+CR and RT groups. Although doing so prevents us from drawing conclusions regarding intent-to-treat effects of randomization to RT+CR or RT alone, we feel that the signal observed over time and between groups is strong enough to warrant replication in an appropriately powered RCT. Additionally, extrapolation of findings reported here should be limited to similar populations of healthy, but overweight and obese, adults in this age group. Findings may differ in older adults at more advanced ages and who may be frail or have more comorbidities.
In conclusion, despite clinically meaningful weight loss and concurrent favorable shifts in total body and thigh composition with short-term RT+CR, improvements do not appear to be sustained long-term relative to RT alone. Specifically, in the 18 months following the intervention, some amount of weight regain occurred in all RT+CR participants, reflected in fat mass and volume accretion, but thigh muscle volume loss. Replication of these findings are necessary and better understanding of the health correlates of weight regain in this population is warranted. Nonetheless, efforts to identify longterm weight loss maintenance strategies to prevent the predisposition for preferential total and regional fat accumulation with weight regain following intentional weight loss in older adults appear prudent.
[fig] Figure 1: (a) Change in total fat and lean mass during and after the intervention period (RT+CR group, only). (b) Change in total thigh muscle and fat volume during and after the intervention period (RT +CR group, only). (c) Change in thigh subcutaneous and intermuscular fat volume during and after the intervention period (RT +CR group, only). [/fig]
[table] Table 2: Thigh composition variables by group and time point [/table]
|
A 22-year-old man with pleural tuberculosis associated hydropneumothorax: Case report and literature review
a b s t r a c tA 22-year-old Asian male presented with fever, non-productive cough, right-sided pleuritic chest pain and was found to have a large right hydropneumothorax. A chest tube was placed. Pleural fluid analysis revealed a lymphocytic predominant exudate and he was subsequently started on four-drug daily antituberculosis therapy (isoniazid, ethambutol, rifampin, pyrazinamide). Pleural biopsy revealed acid-fast bacilli. Given his persistent pleural effusion, he was given four doses of intrapleural tissue plasminogen activator (tPA) and dornase alpha (DNase) via his chest tube over a period of 6 days resulting in clinical and radiologic improvement. Pleural biopsy and pleural fluid culture specimens later revealed Mycobacterium tuberculosis. Intrapleural tPA-DNase therapy has demonstrated improved resolution of infections and shortened hospitalizations for parapneumonic infectious effusions. However, there is little literature on the use of intrapleural fibrinolytics specifically for pleural tuberculosis associated effusions. Furthermore, the American Thoracic Society does not comment on therapeutic thoracentesis or intrapleural fibrinolytic therapy in their recommendations for treatment of pleural tuberculosis. In our case of pleural TB-associated hydropneumothorax, the use of intrapleural tPA-DNase therapy facilitated pleural fluid drainage and resulted in near-complete resolution of the effusion.
# Introduction
Pleural tuberculosis (TB) is rare. The most recent epidemiologic study in 2003 reported 14,874 US TB cases of which 20.4% were exclusively extra-pulmonary. Of these patients, [bib_ref] Controlled trial of intrapleural streptokinase for pleural infection, Maskell [/bib_ref].5% reported pleura as the major site of disease. From 1993 to 2003, pleural TB comprised only 3.9% of all US TB cases [bib_ref] Pleural tuberculosis in the United States: incidence and drug resistance, Baumann [/bib_ref]. Globally there is significant geographical variation, for example the incidence in Spain is nearly six times that of the US [bib_ref] Diagnosis and treatment of tuberculous pleural effusion in, Gopi [/bib_ref].
Little published data exists about the management of pleural TB associated effusions. We report a case of an Asian male with a large hydropneumothorax found to have acid-fast bacilli (AFB) on pleural biopsy and ultimately Mycobacterium tuberculosis in pleural biopsy and fluid cultures.
## Case report
A 22 year-old Asian male smoker presented with ten days of non-productive cough, subjective fevers, night sweats, right-sided pleuritic chest pain and increasing dyspnea. One week prior to presentation, he was empirically treated for pneumonia with levofloxacin without improvement. The patient was an exchange student from China whose last visit home was two months prior to presentation to our facility. Both of the patient's parents had been diagnosed with TB before his birth with unclear treatment status.
The patient's physical exam was remarkable for decreased rightsided breath sounds. Vital signs were: temperature, 98.4 F; heart rate, 102/min, regular; BP, 102/58 mm Hg; respiratory rate, 28/min; oxygen saturation, 95% on room air.
Chest radiograph (CXR) showed a large right hydropneumothorax and right upper lobe infiltrate with scarring [fig_ref] Figure 1: Initial chest radiographs of patient upon presentation [/fig_ref].
Abbreviations: TB, tuberculosis; CT, chest tube; AFB, acid fast bacilli; HD, hospital day; CXR, chest radiograph; ADA, adenosine deaminase; INH, isoniazid; EMB, ethambutol; RIF, rifampin; PZA, pyrazinamide.
A right-sided 32 French chest tube (CT) was placed in the emergency department [fig_ref] Figure 2: Chest radiographs showing resolution of the hydropneumothorax [/fig_ref]. A chest computed tomography scan showed tree-in-bud nodularity in the right upper lobe, possible consolidation in the right lower lobe and calcified hilar lymphadenopathy. Additionally, there were foci of gas in the right-sided hydropneumothorax which could have represented a loculated component. Due to imaging findings suggesting prior granulomatous infection and patient history, there was high suspicion for TB and on airborne precautions were instituted.
A pleural drainage catheter was placed into the hydropneumothorax. Analysis of the pleural fluid obtained from this procedure revealed: pH, 7.70; glucose, 11 mg/dL; protein, 4.5 g/L; LDH, 1458 units/L; WBC, 1340 cells/mL (64% lymphocytes); RBC, 4000 cells/mL; adenosine deaminase (ADA), 57.5 units/L. Additionally, serum protein was 5.6 g/L.
Throughout his hospitalization, the patient had multiple AFB smears and cultures performed (2 concentrated sputum, 2 pleural fluid, 1 bronchoalveolar lavage), which were negative. A pleural biopsy demonstrated 1þ AFB on smear and necrotizing granulomatous inflammation. The patient was started on an antituberculosis regimen (isoniazid (INH) 300 mg, ethambutol (EMB) 800 mg, rifampin (RIF) 450 mg and pyrazinamide (PZA) 1000 mg) because of his findings of a lymphocyte predominant exudative pleural effusion, AFB positive pleural biopsy with necrotizing granulomatous inflammation and history of exposure to TB.
Given the persistent pleural effusion, he was given four doses of tissue plasminogen activator (tPA) and dornase alfa (DNase) instilled via the chest tube over a period of 6 days resulting in improved drainage and clinical and radiologic improvement [fig_ref] Table 1: Dosages of intrapleural fibrinolytic therapy and resultant chest tube outputs [/fig_ref] , [fig_ref] Figure 2: Chest radiographs showing resolution of the hydropneumothorax [/fig_ref]. Fibrinolytic therapy was terminated at the point of near complete resolution of the pleural effusion as evidenced by CXR. The chest tube was removed and the patient was discharged on hospital day 13. Two weeks after discharge, a follow-up CXR demonstrated a persistent small right-sided effusion, however; the patient was asymptomatic [fig_ref] Figure 3: Chest radiographs at patient's two-week follow-up visit [/fig_ref]. Ultimately, the pleural fluid and biopsy cultures were positive for Mycobacterium tuberculosis.
# Discussion
Tuberculous pleuritis should be considered in patients with a lymphocyte-predominant exudative effusion. The etiology of an effusion can be difficult to classify and often requires both pleural fluid analysis and clinical correlation [bib_ref] Pleural effusions, Light [/bib_ref]. Our patient had a pleural fluid protein/serum protein ratio of 0.80 and elevated pleural fluid LDH indicating an exudate. Low pleural fluid glucose (40 mg/dL) narrowed our differential to rheumatoid pleurisy, empyema, malignancy, lupus pleuritis or tuberculous pleurisy [bib_ref] In defense of low glucose level in pleural fluid, Sahn [/bib_ref]. In this patient, serum protein >4 g/dL made tuberculous pleural effusion more likely [bib_ref] Pleural effusions: the diagnostic separation of transudates and exudates, Light [/bib_ref]. Our patient's pleural fluid did not show the typical pH of <7.30 as seen in most tuberculous effusions [bib_ref] The diagnostic value of pleural fluid pH, Good [/bib_ref]. In one metaanalysis, the average pH for tuberculous effusions was 7.18 [bib_ref] Pleural fluid pH: diagnostic, therapeutic, and prognostic value, Houston [/bib_ref]. The pleural fluid ADA value was found to be elevated at 57.5 U/L. The ADA assay has a sensitivity of 92% and a specificity of 90 for diagnosing TB pleurisy [bib_ref] Diagnostic accuracy of adenosine deaminase in tuberculous pleurisy: a meta-analysis, Liang [/bib_ref].
The best diagnostic approach to suspected tuberculous pleuritis is debated. The yield of pleural fluid smears is 10% and of pleural fluid cultures is 25e85% [bib_ref] Pleural tuberculosis in the United States: incidence and drug resistance, Baumann [/bib_ref]. Pleural biopsy histopathology of granulomas or (þ) culture has diagnostic yields ranging from 55 to 93% [bib_ref] Pleural tuberculosis in the United States: incidence and drug resistance, Baumann [/bib_ref] [bib_ref] Thoracoscopic pleural biopsy for tuberculous pleurisy under local anesthesia, Sakuraba [/bib_ref]. Evaluation for TB with thoracentesis plus closed pleural biopsy has a 95% sensitivity, which is roughly on par with thoracoscopy [bib_ref] Closed needle biopsy of the pleura is a valuable diagnostic procedure, Baumann [/bib_ref]. Diagnostic modalities with highest yield are slow to show results making acute workup difficult.
The American Thoracic Society recommends a 6-month regimen for treatment of pleural TB consisting of a 2-month period of INH, [bib_ref] Corticosteroids in the treatment of tuberculous pleurisy. A double-blind, placebo-controlled, randomized study, Wyser [/bib_ref]. Therapeutic thoracentesis is not discussed in the guidelines and is controversial, however it is usually performed if a patient is more than mildly symptomatic [bib_ref] Update on tuberculous pleural effusion, Light [/bib_ref]. Significantly less residual pleural thickening and accelerated recovery of pulmonary function has been seen in patients who received therapeutic thoracentesis for TB associated effusions [bib_ref] Role of therapeutic thoracentesis in tuberculous pleural effusion, Bhuniya [/bib_ref] [bib_ref] Early effective drainage in the treatment of loculated tuberculous pleurisy, Chung [/bib_ref]. However, other studies have concluded that residual pleural thickening is not influenced by this intervention [bib_ref] Pigtail drainage in the treatment of tuberculous pleural effusions: a randomised study, Lai [/bib_ref]. Other drawbacks to thoracentesis include risk of transmission to health providers, bleeding, and pulmonary injury. In our patient's clinical situation a chest tube was placed on presentation for symptomatic relief. Treatment of parapneumonic infectious effusions is comprised of antibiotic therapy and early pleural drainage [bib_ref] Causes and management of common benign pleural effusions, Thomas [/bib_ref]. A small percentage of these patients require additional therapy. The MIST-1 (Multi-Center Intrapleural Sepsis Trial) did not demonstrate any improvement in patient mortality, rate of surgery, or the length of the hospital stay with use of intrapleural tPA therapy alone [bib_ref] Controlled trial of intrapleural streptokinase for pleural infection, Maskell [/bib_ref]. However, the use of intrapleural tPA-DNase therapy in the MIST-2 trial resulted in improved resolution of infection, shortened hospitalization, and resolved effusion in 95% of patients without the need for surgery [bib_ref] Intrapleural use of tissue plasminogen activator and DNase in pleural infection, Rahman [/bib_ref]. Unfortunately, none of the patients in the MIST-1 and MIST-2 trials had isolates of TB recovered making applicability to our case difficult.
A literature review on use of intrapleural tPA-DNase therapy in tuberculous pleural effusions has shown that this therapy improves fluid drainage and lessens residual pleural thickening [fig_ref] Table 2: Studies utilizing intrapleural infusions in treatment of tuberculous pleural effusions [/fig_ref]. However, there is little literature on the use of intrapleural fibrinolytics specifically for pleural TB associated effusions. In our case of pleural TB-associated hydropneumothorax, the use of intrapleural tPA-DNase therapy facilitated pleural fluid drainage and resulted in near-complete resolution of the effusion. Further studies examining the impact of intrapleural lytic therapy on cost, length of stay and patient outcomes in tuberculous effusions are necessary.
# Disclosure
The following authors have no financial or nonfinancial disclosures: Lauren A. Sharan, Thea P. Price, Boyd Hehn, David Manoff, Scott W. Cowan.
[fig] Figure 1: Initial chest radiographs of patient upon presentation. A, Posteroanterior view showing large right hydropneumothorax. B, Lateral view showing hydropneumothorax. [/fig]
[fig] Figure 2: Chest radiographs showing resolution of the hydropneumothorax. A, Following chest tube placement. B, Day 4 after chest tube placement. [/fig]
[fig] Figure 3: Chest radiographs at patient's two-week follow-up visit. A, Posteroanterior view. B, Lateral view. [/fig]
[table] Table 1: Dosages of intrapleural fibrinolytic therapy and resultant chest tube outputs. [/table]
[table] Table 2: Studies utilizing intrapleural infusions in treatment of tuberculous pleural effusions. [/table]
|
Serotonin receptors in depression: from A to B
The role of serotonin in major depressive disorder (MDD) is the focus of accumulating clinical and preclinical research. The results of these studies reflect the complexity of serotonin signaling through many receptors, in a large number of brain regions, and throughout the lifespan. The role of the serotonin transporter in MDD has been highlighted in gene by environment association studies as well as its role as a critical player in the mechanism of the most effective antidepressant treatments -selective serotonin reuptake inhibitors. While the majority of the 15 known receptors for serotonin have been implicated in depression or depressive-like behavior, the serotonin 1A (5-HT ) and 1B (5-HT ) receptors are among the most studied. Human brain imaging and genetic studies point to the involvement of 5-HT and 5-HT receptors in MDD and the response to antidepressant treatment. In rodents, the availability of tissue-specific and inducible knockout mouse lines has made possible the identification of the involvement of 5-HT and 5-HT receptors throughout development and in a cell-type specific manner. This, and other preclinical pharmacology work, shows that autoreceptor and heteroreceptor populations of these receptors have divergent roles in modulating depression-related behavior as well as responses to antidepressants and also have different functions during early postnatal development compared to during adulthood.
## 1
1,2
# Introduction
The serotonin hypothesis of depression has dominated the field of depression for over four decades 1 . This theory is centered on the idea that reduced serotonin signaling is a risk factor in the etiology and/or pathophysiology of major depressive disorder (MDD) 2 . However, the most robust body of evidence for the role of serotonin in depression is the efficacy of increasing extracellular serotonin for the treatment of depression. The discovery that the efficacy of tricyclic antidepressants (TCAs) and monoamine oxidase inhibitor (MAOI) antidepressants was largely due to their serotonergic actions, which prompted the use of serotonin selective reuptake inhibitors (SSRIs), the first among them fluoxetine, to treat depression 3-6 . These drugs act at the serotonin transporter (5-HTT, also known as SERT) and cause increases in extracellular serotonin, which is the purported mechanism of action 6-8 . Many subsequent drugs inhibiting serotonin reuptake have shown behavioral efficacy as antidepressant drugs, suggesting that increasing synaptic serotonin levels may lead to the treatment of depression 6,9 .
Despite the relative success in treating depression by increasing extracellular serotonin, there is a lack of strong evidence supporting a direct correlation between low serotonin signaling and depression. While some studies report an association between levels of platelet serotonin and depression, this has not been a consistent finding in large sample sets, and it is also unclear how platelet levels are related to brain levels of serotonin . Additionally, few studies report direct correlations between cerebrospinal fluid 5-hydroxyindoleacetic acid (5-HIAA), a serotonin metabolite, and depression . Low levels of tryptophan have been consistently linked to depression; however, these effects could be independent of serotonin . The lack of consistent clear-cut abnormalities in global measures of serotonin signaling isn't surprising if one considers the complexity of the receptors at which serotonin binds, the intricate neuroanatomical circuitry of the serotonin system, and the developmental role serotonin plays as a neurotrophic factor . Many recent studies have focused on understanding the mechanisms through which serotonin affects depression by studying the impact of 5-HTT and the 15 known receptors through gene-association studies, human brain imaging, and pharmacological and genetic mouse models 19 .
The success in treating depression by targeting the transporter with SSRIs prompted investigations into whether variability in 5-HTT expression levels could be involved in the etiology of depression. A highly cited study showed that there is an association between a polymorphism in the serotonin transporter (5-HTTLPR) and susceptibility to developing depression 20 . This and other studies have shown that the short "s" allele, which results in lower levels of 5-HTT expression (at least in vitro) and therefore increased extracellular 5-HT, is associated with a higher risk of depression when combined with stressful life events 21,22 . This discovery would be unexpected if developmental considerations were not considered. Although inhibiting the function of the transporter during adulthood decreases depressive symptoms as in the case of SSRIs, reduced expression of 5-HTT during development may increase depressive behavior in adulthood. A human functional magnetic resonance imaging (fMRI) study supports this,
showing that short allele carriers show morphological and functional alterations in limbic circuits 23 . Additionally, mice lacking 5-HTT throughout life display increased depressive-like behaviors, and pharmacological blockade of 5-HTT in mice exclusively during early postnatal development resulted in increased adult depressive behavior 24 . These results highlight the differences in developmental versus adult effects of altered serotonin neurotransmission on depression.
In addition to the serotonin transporter, the majority of the 15 serotonin receptors have been implicated in the modulation of depression, depressive-like behaviors, or the response to antidepressant treatment 19 . There are numerous pre-clinical studies which have investigated the role of serotonin receptors using pharmacological manipulations and genetic knockout (KO) models in rodents [fig_ref] Table 1: Preclinical evidence supporting the role for serotonin receptors in depression [/fig_ref]. Given the breadth of this literature, this review will focus on two receptors that are among the most extensively studied for their role in modulating depression, the 5-HT 1A and 5-HT 1B receptor subtypes. In addition, attention will be paid to population-dependent and development-dependent effects of serotonin signaling at these receptors and will draw from both rodent and human studies.
The 5-HT 1A and 5-HT 1B receptors are both inhibitory Gi/o-coupled seven transmembrane receptors that are located throughout the brain 25-27 . A major difference between these two receptors is their subcellular distribution 28 . 5-HT 1A receptors are somatodendritic, while 5-HT 1B receptors are located on axon terminals 27,29 . This difference is also reflected in their mechanisms of inhibitory action [fig_ref] Figure 1: Schematic illustrating the inhibitory effects of serotonin [/fig_ref]. Activation of either receptor causes decreased neurotransmitter release; however, 5-HT 1A receptor activation causes hyperpolarization, leading to decreased firing, while 5-HT 1B receptors inhibit voltage-gated calcium channels in the presynaptic terminal 30-32 . Another mechanism for 5-HT 1B receptormediated inhibition is via effects on 5-HTT, and activation of the 5-HT 1B receptor increases serotonin reuptake 33,34 .
Both 5-HT 1A and 5-HT 1B receptors act as autoreceptors located on serotonin neurons and also have heteroreceptor populations located on non-serotonin receptors [fig_ref] Figure 2: Diagram summarizing the roles of autoreceptor and heteroreceptor populations of serotonin [/fig_ref]. Although the mRNA in the raphe (corresponding to autoreceptors) is comparable between the two receptors, their heteroreceptors have distinct patterns of expression 35 . 5-HT 1A receptors are enriched in the hippocampus and cortex, while 5-HT 1B receptors are highly expressed in the basal ganglia 36,37 . These differences in mechanism of action and localization may play a role in the different functional effects of these receptors.
While this review focuses on the contribution of 5-HT 1A and 5-HT 1B receptors in depression and depressive-like behaviors, these receptors also modulate other psychiatric-relevant phenotypes. For example, alterations in 5-HT 1A receptor expression influence anxiety behavior, and 5-HT 1B receptor signaling affects reward-and impulsivity-related phenotypes. These receptor-based differences in serotonergic regulation of emotional behavior, which segment into endophenotypes, could contribute to the heterogeneity of symptoms found in MDD 38 . Understanding the neural circuits that subserve these receptor-based and endophenotype-based Antagonists have pro-cognitive effects 160 *Number of PubMed hits based on the search terms including "depression" and the receptor as of August 25, 2016.
N.B. 5-HT1D, 1E, 1F, 3B, and 5B are not included in the chart owing to a lack of published research concerning the role of these receptors in behavior.
5-HT, serotonin; KO, knockout; SSRI, selective serotonin reuptake inhibitor.
differences can help clarify the often confusing and sometimes contradictory findings from various preclinical approaches. From a behavioral perspective, these phenotypes can be segmented through formal unsupervised factor analyses to better divide depressive behaviors into meaningful endophenotypes. Then predictors of the different endophenotypes could be tested by including genetic or pharmacological manipulations.
## 5-ht 1a and depression
Of the 15 known serotonin receptors, the 5-HT 1A receptor is the most studied for its role in depression 39 . Quantification of 5-HT 1A receptor levels in humans from post mortem and positron emission tomography (PET) imaging studies reveals an increased level of 5-HT 1A receptors in patients diagnosed with MDD 40-42 . Gene association studies have linked a polymorphism in the 5-HT 1A regulatory region (rs6295; G-1019C) with receptor levels in the brain and also to increased risk for depression . The GG genotype at this single nucleotide polymorphism (SNP) is associated with altered levels of 5-HT 1A receptor expression and reduced responsiveness to antidepressant treatment . Additionally, clinical studies have revealed antidepressant effects of buspirone and other 5-HT 1A receptor agonists 49,50 .
Rodent models have also shown that 5-HT 1A receptor agonists, such as 8-hydroxy-2-(di-n-propylamino) tetralin (8-OH-DPAT), can have acute antidepressant-like effects 51-53 . These effects are blocked by 5-HT 1A receptor antagonists, suggesting that the antidepressant-like response is specific to 5-HT 1A receptor signaling 54 . 5-HT 1A heteroreceptors, expressed throughout the limbic system, are the likely site of action for these acute 5-HT 1A receptor-mediated effects 50,55 . On the other hand, 5-HT 1A autoreceptors work in opposition to the heteroreceptors, leading to pro-depressive effects. Specifically, activation results in hyperpolarization and reduced firing of raphe neurons, leading to diminished serotonin release in projection regions 56 . Therefore, stimulation of 5-HT 1A autoreceptors from increased extracellular serotonin following SSRI treatment is thought to oppose SSRI actions by downregulating serotonin neuron activity 57 . Over the first few weeks of treatment, these receptors desensitize, which may underlie the delayed behavioral efficacy of SSRIs 58 . Therefore, blocking 5-HT 1A autoreceptor activation has been introduced as an adjunctive therapy to SSRIs. 5-HT 1A receptor partial agonists such as pindolol, and more recently vilazodone, have been shown to be an effective adjunctive therapy to SSRIs in clinical studies 59-62 . The development of new agonists that preferentially target subpopulations of 5-HT 1A receptors, for example autoreceptors versus heteroreceptors, potentially through biased agonism, may be useful tools for the treatment of MDD 63 .
Differences in receptor levels have also been modeled in mice by using genetic loss-of-function models and have allowed causal links between receptor expression levels and depressive-like behavior. 5-HT 1A receptor KO mice have an anti-depressive phenotype 64,65 . Tissue-specific KOs have been especially valuable for the dissection of this phenotype and have allowed investigations into the distinct roles of different populations of receptors 66 . The absence of heteroreceptors results in increased depressive-like behavior-as measured in the forced swim test. This mouse model also allowed for temporal control of receptor expression, which revealed a developmental sensitive period for the effect of heteroreceptors on depressive-like behavior. Specifically, knockdown of 5-HT 1A heteroreceptors in adulthood was not sufficient to produce the depressive-like behavior. On the other hand, reduction of autoreceptors in adulthood increased mobility in the forced swim test, suggesting an "anti-depressed" phenotype.
Preclinical studies have also confirmed a causal role for alterations in 5-HT 1A receptor expression in antidepressant efficacy. 5-HT 1A receptor KO mice do not show a behavioral response to fluoxetine 67 . As expected from the pharmacology work, this effect is not mediated by autoreceptors, since reduced expression of 5-HT 1A autoreceptors actually increases the speed and efficacy of SSRI response, requiring only 8 days of fluoxetine treatment to show a behavioral antidepressant-like response 68 . Recent data show that 5-HT 1A heteroreceptors are critical for an effective behavioral response to an SSRI in mice 69 . Genetic or viral deletion of 5-HT 1A receptors specifically in the dentate gyrus of the hippocampus reduced the behavioral response to fluoxetine. Furthermore, expression of 5-HT 1A receptors only in the dentate gyrus was sufficient for normal antidepressant-like responses. These results importantly demonstrate a mechanism for 5-HT 1A -mediated antidepressant effects localized in the mature granule cells of the dentate gyrus.
## 5-ht 1a and other psychiatric-relevant phenotypes
Anxiety behavior is also modulated strongly by the 5-HT 1A receptor, and, among depressed patients, almost half have a comorbid anxiety disorder 70 . In preclinical studies, 5-HT 1A receptor agonists have anxiolytic effects, and 5-HT 1A receptor KO mice display increased anxiety-like behavior . The effect has a developmental sensitive period, since early developmental but not adult rescue of the receptor was sufficient to restore the normal phenotype in the KO 73 . Consistent with this, early postnatal blockade of 5-HT 1A receptors, through genetic or pharmacological methods, also leads to increased anxiety . Recent work has shown that the sensitive period is peri-pubertal, and tissue-specific KO mice point to a role for autoreceptors during this period of development .
Other psychiatric disorders have also been linked to the 5-HT 1A receptor, including bipolar disorder and post-traumatic stress disorder 78,79 . Additionally, the SNP rs6295 found in the premotor region that is associated with risk for depression is also linked with psychiatric hospitalization, a history of substance abuse, and prior suicide attempts 43 . Consistent with the studies in depression, the G allele is associated with reduced expression of the 5-HT 1A receptor in the prefrontal cortex and an increased risk for psychiatric outcomes. Interestingly, the effects on receptor expression were also seen in the brain during early human embryonic development, suggesting its potential importance in mediating developmental contributions to adult depression. Finally, there were associations with childhood maltreatment with trends towards significant genotype by environment interactions 40 .
## 5-ht 1b and depression
While the 5-HT 1B receptor is best known for its role in regulating aggressive and impulsive behavior, it also plays an important role in modulating depression. Activation of the 5-HT 1B receptor decreases serotonin levels in the brain through effects on release, synthesis, and reuptake . In humans, reduced 5-HT 1B receptor function is associated with MDD 82 . Additionally, patients with MDD are less responsive to 5-HT 1B receptor agonists, suggesting reduced expression or desensitization . This is consistent with clinical studies showing that 5-HT 1B receptor agonists produce antidepressant effects in humans 85-87 . This has also been shown in mice, with specific agonists resulting in antidepressant-like behavior 88,89 . However, genetic KO of the receptor also results in antidepressant-like behavior, suggesting that this is possibly caused by compensatory effects 90-94 .
Both autoreceptor and heteroreceptor populations of 5-HT 1B receptors have been implicated in depressive-like behaviors using rodent models. However, since 5-HT 1B receptors are located on presynaptic terminals, heteroreceptors and autoreceptors have overlapping localization 95 . This rules out brain imaging and pharmacological manipulations in preclinical models as tools to differentiate the role of the two populations of receptors. Therefore, it has been only the recent availability of a tissue-specific genetic mouse model that has allowed the dissection of the role of 5-HT 1B receptors in the regulation of behavior 96 .
Our recent studies show that selective ablation of 5-HT 1B autoreceptors results in decreased depressive-like behaviors in mice 97 . These mice also show increased elevations in serotonin levels compared to controls following SSRI administration, suggesting a potential mechanism of action for the behavioral effects. Specifically, removing the terminal auto-inhibition may result in increased serotonin in projection regions that are relevant to depressive behavior. Furthermore, we also showed that the impact of 5-HT 1B autoreceptors on behavior was not due to developmental expression, since the phenotype was not recapitulated in a mouse with developmental knockdown. These data are consistent with other evidence suggesting a pro-depressive role for the activation of 5-HT 1B autoreceptors 98,99 . For example, 5-HT 1B mRNA is elevated in the raphe of rats following stress and in models of depression such as learned helplessness, and viral overexpression of 5-HT 1B receptors in the raphe results in depressive-like behavior following stress 100 . In rats, reductions in 5-HT 1B receptor mRNA in the raphe are seen following SSRI treatment in post mortem brains 101,102 . This effect isn't seen in other brain regions such as the cortex, hippocampus, or striatum, suggesting that this effect is specific to autoreceptors. Additionally, another study showed that 5-HT 1B autoreceptors may desensitize following SSRI treatment, similar to 5-HT 1A autoreceptors 103 . Finally, a recent PET study in humans reported that following effective cognitive behavioral therapy for depression, 5-HT 1B receptor binding was reduced in the brainstem 104 .
There is evidence which suggests an opposing role for 5-HT 1B heteroreceptors in depressive behaviors. Activation of 5-HT 1B heteroreceptors in a rodent serotonin depletion model (to remove the contribution of autoreceptors) results in an antidepressant-like effect 105 . Additionally, reduced expression of 5-HT 1B heteroreceptors in the ventral striatum is associated with depression in humans 82 . Finally, 5-HT 1B receptors located in the ventral striatum have been suggested to interact with p11 (a 5-HT 1B receptor-binding protein) to affect depression-related behaviors 106,107 .
## 5-ht 1b and other psychiatric-related phenotypes
Reward dysfunction is a major symptom of MDD which is mediated, in part, by altered signaling in the mesolimbic reward system 108-112 . 5-HT 1B receptors have been implicated in the neural basis of dysregulated reward sensitivity in a number of human studies and preclinical models , and both 5-HT 1B receptor protein and mRNA are located within the mesolimbic pathway in the nucleus accumbens (NAc) and ventral tegmental area (VTA) 95 . Additionally, activation of 5-HT 1B receptors in the VTA increases dopamine levels in the NAc, potentially via effects on GABAergic signaling in the VTA 117 .
Many studies linking the receptor to functional deficits in reward processing have focused on addiction. Polymorphisms in the 5-HT 1B receptor gene have also been associated with drug and alcohol abuse 118-120 . Additionally, a PET imaging study revealed increased 5-HT 1B receptor binding in pathological gamblers, who have known deficits in reward sensitivity, and gambling disorder is highly comorbid with depression and alcohol and substance use disorders . Another PET imaging study shows that there is reduced 5-HT 1B receptor binding in cocaine-dependent participants compared to in healthy controls 122 . In preclinical models, 5-HT 1B receptor KO mice are more motivated to self-administer cocaine 123 . Consistent with this, 5-HT 1B receptor agonists attenuate the motivation for cocaine but paradoxically increase the rewarding effects of cocaine 124 . These effects are mediated by 5-HT 1B receptor expression on medium spiny neurons in the NAc, likely through their projections to the VTA 125,126 . Additionally, 5-HT 1B receptors are required for the rewarding properties of social interaction, supporting an impact on general reward systems 114 .
5-HT 1B receptors are also implicated in impulsive aggression. In humans, polymorphisms in the gene encoding 5-HT 1B receptors have been associated with aggression, suicide, and disorders that include impulsivity as a core phenotype, including attention deficit hyperactivity disorder and substance use disorder 115,120,127 . In mice, 5-HT 1B receptor KOs are highly aggressive in tests of male and female aggression and also display increased impulsivity 128-130 . Additionally, 5-HT 1B receptor agonists are known as "serenics" because they decrease aggression 131 . While the aggressive and impulsive phenotype was originally thought to be modulated by the same underlying circuits, our recent work shows that distinct populations of 5-HT 1B receptors modulate aggression and impulsivity 96 . Furthermore, developmental expression of the 5-HT 1B receptor influences aggression, while adult expression modulates impulsive behavior.
# Conclusion
There is a considerable body of research that implicates serotonin in the modulation of depression and depression-related behaviors.
The preclinical work delineating the effects of signaling through the 5-HT 1A and 5-HT 1B receptors has been made possible because of careful pharmacological studies as well as the development of transgenic mouse models that have allowed for tissue-specific and inducible knockdown. These studies have highlighted the complexity of serotonin receptors, showing that their role varies through the lifespan and by cell-type population. Additionally, the availability of specific radioligands for PET imaging of these receptors has allowed for the translation of findings from preclinical work to humans. The large number of studies concerning the role of these receptors is partially due to the fact that the 5-HT 1 receptor subtypes were some of the first discovered, and it may be only a matter of time before the roles of more newly discovered receptors are clarified 17 .
Despite the amassing of evidence of serotonin receptor-specific involvement in depression, the primary pharmaceutical treatment strategy for depression remains the inhibition of serotonin reuptake. The lack of new treatment options is surprising given the need for them, since current SSRI treatments are ineffective in one-third of patients 132 . Additionally, the majority of patients, as seen in the STAR*D study, don't respond to administration of the first SSRI treatment, requiring multi-step treatment plans that take months 132 . Furthermore, the considerable differences in treatment outcome also emphasize the heterogeneity of the depressed patient population. A better understanding of receptor signaling andneural circuit mechanisms by which serotonin affects depression may inform the development of novel, more targeted drugs that influence specific receptors, signaling cascades, or time periods. Also, personalized treatment plans could be developed based on symptoms, biomarkers, or pathophysiological presentation.
during a major depressive episode predicts poor treatment response:
helpless rats in the forced swim test. Synapse. 2004; 52(1): 73-5.
[fig] Figure 1: Schematic illustrating the inhibitory effects of serotonin (5-hydroxytryptamine, 5-HT) 1A (5-HT 1A ) (red) and 5-HT 1B (blue) receptors on the normal firing and neurotransmitter release of a neuron (top). Activation of 5-HT 1A receptors results in decreased firing (middle), while activation of 5-HT 1B receptors causes decreased neurotransmitter release through actions in the presynaptic terminal (bottom). [/fig]
[fig] Figure 2: Diagram summarizing the roles of autoreceptor and heteroreceptor populations of serotonin (5-hydroxytryptamine, 5-HT) 1A (5-HT 1A ) and 5-HT 1B receptors on behavior during development and adulthood. 5-HTT, serotonin transporter. [/fig]
[table] Table 1: Preclinical evidence supporting the role for serotonin receptors in depression. [/table]
|
Acute zonal occult outer retinopathy: optical coherence tomography angiography findings and treatment response
Acute zonal occult outer retinopathy (AZOOR) is a rare condition primarily affecting the outer retina based on electrophysiologic studies.
# Introduction
Acute zonal occult outer retinopathy (AZOOR), first reported by Gass in 1992 [bib_ref] Acute zonal occult outer retinopathy. Donders Lecture: The Netherlands Ophthalmological Society, Gass [/bib_ref] , typically affects young adult women who present with central vision loss with or without photopsia due to rapid loss of one or more large zones of outer retinal function. The exact pathogenesis of the disease is not well known. Infectious, autoimmune, or inflammatory etiologies were supposed to be responsible in the emergence of disease [bib_ref] Acute zonal occult outer retinopathy. Donders Lecture: The Netherlands Ophthalmological Society, Gass [/bib_ref] , [bib_ref] White spot syndromes of the retina: a hypothesis based on the common..., Jampol [/bib_ref]. Few studies have described the role of multimodal imaging in AZOOR [bib_ref] Acute zonal occult outer retinopathy: a classification based on multimodal imaging, Mrejen [/bib_ref] ,. A common consensus has not been achieved for treatment in the literature. We have evaluated clinical features of AZOOR in a 64-year-old man using multimodal imaging including optical coherence tomography angiography (OCTA) (AngioVue, RTVue XR Avanti; Optovue, Inc., Freemont, CA, USA) and treatment response to combined systemic antiviral and immunsupressive drugs during follow-up.
## Case description
A 64-year-old man presented with sudden onset decrease in vision and photopsia in the right eye since 10 days. Systemic and ocular history was normal. There was no viral infection history preceding the symptoms. The best corrected visual acuity (BCVA) in the patient's right eye was counting fingers at 2 m and 9/10 in his left eye. There was nuclear sclerosis in both eyes. Fundus examination demonstrated an annular ring of yellowish reflex with an about 4-5 disc diameter at the posterior pole in the right eye . Anterior uveitis or vitritis was not detected. There was slightly increased hyperautofluorescence, more prominent at the temporal border on fundus autofluorescence (FAF) imaging . Infrared fundus image showed the lesion contour consistent with the FAF and color images [fig_ref] Figure 3: Infrared [/fig_ref]. Spectral domain optical coherence tomography (SD-OCT) revealed hyperreflective dots between the retinal pigment epithelium (RPE) and the ellipsoid zone (EZ) and loss or irregularity of the EZ [fig_ref] Figure 4: SD-OCT images at the time of diagnosis [/fig_ref]. En-face OCTA outer retinal slab demonstrated multiple hyperreflective dots in a starry-sky appearance . The patient was diagnosed as AZOOR and started on oral valgancyclovir (1 g three times daily for a week, once a day for 3 weeks), oral azathioprine (50 mg three times daily) and a weekly tapering course of oral prednisolone (starting at 1 mg/kg body weight). At 2-week examination, significant visual and anatomical improvements were observed. BCVA in the right eye improved to 6/10. The retinal lesion with annular ring of yellowish reflex disappeared . There was a marked recovery in the EZ and the outer segment of the photoreceptors on SD-OCT images [fig_ref] Figure 8: SD-OCT images at 2-week examination after the treatment [/fig_ref]. En-face structural OCTA images showed an obvious decrease in the hyperreflective dots [fig_ref] Figure 9: OCTA images at 2-week examination after the treatment [/fig_ref]. The patient was kept under the same treatment and the follow-up continued. The left eye was not affected during the follow-up.
# Discussion
Although AZOOR was first described in 1992 [bib_ref] Acute zonal occult outer retinopathy. Donders Lecture: The Netherlands Ophthalmological Society, Gass [/bib_ref] , its etiopathogenesis and treatment remain unclear. Based on the electrophysiologic studies, the primary lesions in AZOOR were suggested as the dysfunction and/or degeneration of the photoreceptor outer segment [bib_ref] Loss of photoreceptor outer segment in acute zonal occult outer retinopathy, Li [/bib_ref]. The fundus view may be normal or damaged at different levels, depending on the stage of the disease and the location of the lesion [bib_ref] Zonal occult outer retinopathy, Rodriguez-Coleman [/bib_ref] , [bib_ref] The retinal features of acute zonal occult outer retinopathy (AZOOR), Smetana [/bib_ref] , [bib_ref] Collateral damage in acute zonal occult outer retinopathy, Spaide [/bib_ref]. However, some features were reported as characteristic, including a demarcating line of the progression at the level of the outer retina and a trizonal pattern of sequential involvement of the outer retina, retinal pigment epithelium, and choroid, as well as frequent zonal progression. AZOOR can manifest as unilateral and asymmetric lesions that have a delineating line to segregate the normal fundus from the AZOOR lesion. Supporting the characteristic color fundus images with SD-OCT and FAF images is very important in the diagnosis of AZOOR. OCTA is a non-invasive, innovative technology for imaging the microvasculature of the retina and the choroid. OCTA has been reported to be useful in the diagnosis and understanding of many retinal conditions [bib_ref] Quantification of Diabetic Macular Ischemia Using Optical Coherence Tomography Angiography and Its..., Samara [/bib_ref].
There are a few reports on OCTA in AZOOR disease in the literature. In the first report by , OCTA was used only to confirm the presence of choroid neovascular membrane in AZOOR. In another case reported by Naik et al. in 2018 [bib_ref] Acute zonal occult outer retinopathy: Is optical coherence tomography angiography useful?, Naik [/bib_ref] , the structural en-face OCTA images demonstrated the hyper-reflective dot structures leading to a starry-sky view at presentation. The authors postulated these hyper-reflective dots as degenerating photoreceptor segments. Mehrotra et al. [bib_ref] Panoramic optical coherence tomography angiography features in acute zonal occult outer retinopathy, Mehrotra [/bib_ref] presented an AZOOR case with an orange-yellow-colored demarcation line in the peripapillary area. In that case, the OCTA panorama showed an increased decorrelation signal at the deeper capillary plexus along with the projection artifacts of the superficial capillary plexus. The authors have postulated that an OCTA panorama centered on the fovea with a wider field of view can prove as a potential imaging marker and can subserve as an additional tool to FAF to accurately diagnose and manage AZOOR. In our case, the OCTA findings were in agreement with the report by Naik et al. [bib_ref] Acute zonal occult outer retinopathy: Is optical coherence tomography angiography useful?, Naik [/bib_ref]. Hyperreflective dots in a starry-sky appearance on en-face OCTA outer retinal slab may be characteristic when evaluated with other typical findings both in diagnosis and treatment response.
There are some treatment approaches including observation without any treatment, immunosuppressive agents, systemic or intravitreal steroids, and antiviral therapy, and the results were reported in contradiction to one another [bib_ref] Zonal occult outer retinopathy, Rodriguez-Coleman [/bib_ref] , [bib_ref] Acute zonal occult outer retinopathy: a long-term follow-up study, Gass [/bib_ref] ,, [bib_ref] Improvement of central visual function following steroid pulse therapy in acute zonal..., Kitakawa [/bib_ref] , [bib_ref] Systemic corticosteroids therapy in the management of acute zonal occult outer retinopathy, Chen [/bib_ref] ,, [bib_ref] Patients with an acute zonal occult outer retinopathy-like illness rapidly improve with..., Mahajan [/bib_ref]. In a study of 51 patients reported by Gass et al. [bib_ref] Acute zonal occult outer retinopathy: a long-term follow-up study, Gass [/bib_ref] , systemic corticosteroids were used in 39 of 113 instances of acute visual loss. The authors reported no difference between 3/8 GMS Ophthalmology Cases 2022, Vol. 12, ISSN 2193-1496 treated and non-treated patients. In the same study, systemic steroid and acyclovir or valacyclovir were prescribed for 11 episodes without apparent therapeutic benefit. The authors concluded that there was no treatment of proven value. In a case report by Hoang et al., treatment with antiviral and immunomodulatory drugs failed to control the disease progression. Rodriguez-Coleman et al. [bib_ref] Zonal occult outer retinopathy, Rodriguez-Coleman [/bib_ref] reported a case that was given a treatment trial of systemic steroids and acyclovir with no clinical benefit. In contrast, Kitakawa et al. [bib_ref] Improvement of central visual function following steroid pulse therapy in acute zonal..., Kitakawa [/bib_ref] reported a case treated with pulse steroids, followed by oral prednisolone tapered over one month, and improvements in visual acuity and imaging findings after fourteen months.
Chen et al. [bib_ref] Systemic corticosteroids therapy in the management of acute zonal occult outer retinopathy, Chen [/bib_ref] indicated beneficial results of systemic steroids in a retrospective study with a mean follow-up of 47 months. In a retrospective case series, disease stability or improvement was reported after intravitreal steroids. It was noted that treatment-related ocular side effects such as cataract, ocular hypertension, and central serous retinopathy were remarkable. Mahajan and Stone [bib_ref] Patients with an acute zonal occult outer retinopathy-like illness rapidly improve with..., Mahajan [/bib_ref] reported 3 cases with a presumptive diagnosis of AZOOR that were responsive to oral valacyclovir without steroid use. Nakao et al. [bib_ref] Spontaneous remission of acute zonal occult outer retinopathy: follow-up using adaptive optics..., Nakao [/bib_ref] presented a case diagnosed with AZOOR with spontaneous remission without any treatment after two months. The predictability of the disease is low due to insufficient data. Referring to recent publications, stabilization has been reported to have occurred within 6 months in the majority of patients [bib_ref] Acute zonal occult outer retinopathy: a long-term follow-up study, Gass [/bib_ref] , [bib_ref] Acute zonal occult outer retinopathy, Monson [/bib_ref]. However, in a small percentage of cases, the disease continued to worsen after 6 months [bib_ref] Acute zonal occult outer retinopathy in Japanese patients: clinical features, visual function,..., Saito [/bib_ref]. In our case, significant improvement in both anatomic and visual function was detected in the second week after combined drug treatment. This improvement continued up to the 2-month examination.
The limitation of the case report is the lack of the anterior chamber or vitreous tap for detecting viral etiology. Although the cause of the disease is unknown, autoimmunity or viral causes are considered in the foreground [bib_ref] Acute zonal occult outer retinopathy: a long-term follow-up study, Gass [/bib_ref] , [bib_ref] one disease or many?, Jampol [/bib_ref]. Gass et al. [bib_ref] Acute zonal occult outer retinopathy: a long-term follow-up study, Gass [/bib_ref] suggested several reasons for viral etiology, including the asymmetrical nature of the disease; the lack of response to the administration of corticosteroids; the lack of correlation between a history of autoimmune conditions, bilateral disease and disease severity; the absence of a family history of AZOOR; and the absence of circulating retinal antibodies in many patients. We also know that many autoimmune disorders tend to affect young women [bib_ref] The Prevalence of Autoimmune Disorders in Women: A Narrative Review, Angum [/bib_ref]. In our case, we thought about a possible viral etiology because of the patient's age and gender, the absence of a systemic history of autoimmune diseases, the asymmetric nature of retinal involvement, and the unilateral presentation of the disease. Because of this, valgancyclovir was added to the prescription. Despite the potential to be bilateral, the second eye was not involved during the follow-up period. The choroid was not affected during follow-up. Both unilateral involvement and limited zonal damage may be associated with early diagnosis and initiation of broadspectrum systemic therapy. Since immunosuppressive agents take many weeks to have an effect, improvement of the clinical findings in the early weeks are most probably due to the systemic steroids. Valgancyclovir may contribute to the absence of a late worsening of the clinical findings at two months. However, this needs to be verified. Surely, a spontaneous remission is also possible. Even if the treatment described has led to improved findings, this does not mean that there is a causal relationship between diagnosis and therapy. Undoubtedly, further studies with a longer follow-up and a larger series are needed to reach a conclusion on etiology and treatment. In addition to other diagnostic tools, OCTA can provide lock tips in diagnosis and follow-up in AZOOR.
[fig] Figure 1, Figure 2: Fundus images at the time of diagnosis (a) Fundus appearance of the AZOOR lesion as an annular ring of yellowish reflex with an about 4-5 disc diameter in the right eye, (b) fundus photo showing the normal left eye FAF images at the time of diagnosis (a) Slightly increased autofluorescence, more prominent at the border on FAF imaging in the right eye, (b) normal autofluorescence image of the left eye [/fig]
[fig] Figure 3: Infrared (IR) images at the time of diagnosis (a) IR image of the right eye showed the lesion contour consistent with the FAF and color images, (b) IR image of the left eye [/fig]
[fig] Figure 4: SD-OCT images at the time of diagnosis (a) Spectral domain optical coherence tomography (SD-OCT) revealed hyperreflective dots between RPE and the ellipsoid zone (EZ) and loss or irregularity of the EZ, (b) SD-OCT image of the left eye [/fig]
[fig] Figure 5, Figure 6, Figure 7: OCTA images at the time of diagnosis (a) Multiple hyperreflective dots in a starry-sky appearance on the en-face OCTA outer retinal slab of the right eye, (b) OCTA image of the left eye Fundus images at 2-week examination after the treatment (a) The retinal lesion with the annular ring of yellowish reflex disappeared on the color fundus image of the right eye, (b) view of left fundus FAF images at 2-week examination after the treatment (a) FAF image of the right eye, (b) FAF image of the left eye [/fig]
[fig] Figure 8: SD-OCT images at 2-week examination after the treatment (a) Marked recovery in EZ and outer segment of photoreceptors on SD-OCT images in the right eye, (b) SD-OCT image of the left eye [/fig]
[fig] Figure 9: OCTA images at 2-week examination after the treatment (a) An obvious decrease in the hyperreflective dots on en-face OCTA, (b) en-face OCTA image of the left eye Notes [/fig]
|
Background: The use of computerized image analysis for the study of nuclear texture features has provided important prognostic information for several neoplasias. Recently fractal characteristics of the chromatin structure in routinely stained smears have shown to be independent prognostic factors in acute leukemia. In the present study we investigated the influence of the fractal dimension (FD) of chromatin on survival of patients with multiple myeloma.Methodology: We analyzed 67 newly diagnosed patients from our Institution treated in the Brazilian Multiple Myeloma Study Group. Diagnostic work-up consisted of peripheral blood counts, bone marrow cytology, bone radiograms, serum biochemistry and cytogenetics. The International Staging System (ISS) was used. In every patient, at least 40 digital nuclear images from diagnostic May-Grü nwald-Giemsa stained bone marrow smears were acquired and transformed into pseudo-3D images. FD was determined by the Minkowski-Bouligand method extended to three dimensions. Goodness-of-fit of FD was estimated by the R 2 values in the log-log plots. The influence of diagnostic features on overall survival was analyzed in Cox regressions. Patients that underwent autologous bone marrow transplantation were censored at the day of transplantation.Principal Findings: Median age was 56 years. According to ISS, 14% of the patients were stage I, 39% were stage II and 47% were stage III. Additional features of a bad prognosis were observed in 46% of the cases. When stratifying for ISS, both FD and its goodness-of-fit were significant prognostic factors in univariate analyses. Patients with higher FD values or lower goodness-of-fit showed a worse outcome. In the multivariate Cox-regression, FD, R 2 , and ISS stage entered the final model, which showed to be stable in a bootstrap resampling study.Conclusions: Fractal characteristics of the chromatin texture in routine cytological preparations revealed relevant prognostic information in patients with multiple myeloma.
# Introduction
Multiple myeloma (MM) is a clonal proliferation of malignant plasma cells characterized by a very heterogeneous disease outcome, varying from relatively asymptomatic slowly progressing forms to a frankly aggressive course. Moreover, important variations in tumor cell biology have been described, thus revealing a heterogeneous disorder [bib_ref] Irregular nuclear shape of bone marrow plasma cells defines a multiple myeloma..., Leleu [/bib_ref]. In 2005 the International Staging System (ISS) was introduced, based on the serum concentrations of ß2-microglobulin and albumin, which provide prognostic stratification of the patients [bib_ref] Irregular nuclear shape of bone marrow plasma cells defines a multiple myeloma..., Leleu [/bib_ref] [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref] [bib_ref] Chromosomal translocations t(4;14), t(11;14) and proliferation rate stratify patients with mature plasma..., Tinguely [/bib_ref] [bib_ref] Identification of new overexpressed genes related to cell proliferation, stimulation and apoptosis..., Ortega [/bib_ref] [bib_ref] Multiple myeloma: new staging systems for diagnosis, prognosis and response evaluation, Rajkumar [/bib_ref]. However, patient's age, degree of anemia, serum creatinine and calcium levels, dehydrogenase activity (LDH) as well as cytogenetic abnormalities have been described as additional prognostic factors [bib_ref] Irregular nuclear shape of bone marrow plasma cells defines a multiple myeloma..., Leleu [/bib_ref] [bib_ref] Chromosomal translocations t(4;14), t(11;14) and proliferation rate stratify patients with mature plasma..., Tinguely [/bib_ref] [bib_ref] Multiple myeloma: new staging systems for diagnosis, prognosis and response evaluation, Rajkumar [/bib_ref]. Plasma cell morphology has also been recognized as an independent prognostic factor for more than two decades [bib_ref] Irregular nuclear shape of bone marrow plasma cells defines a multiple myeloma..., Leleu [/bib_ref] [bib_ref] Mature plasma cells as indicator of better prognosis in multiple myeloma. New..., Goasguen [/bib_ref]. However, the recognition of prognostically relevant cell atypias and maturation of the neoplastic plasma cells is dependent on the morphologist's expertise, thus leading to a considerable interobserver variability. On the other hand, experimental studies using computerized image analysis have shown that subvisible modifications of the chromatin architecture in myeloma cells are associated with important changes of cell physiology, such as acquired drug resistance [bib_ref] Nuclear chromatin patterns in 3 glucocorticoid-resistant RPMI 8226 human myeloma cell sub-lines:..., Genty [/bib_ref]. The quantitative analysis of the distribution pattern of chromatin and other nuclear components has been shown to be of prognostic importance in several neoplasias [bib_ref] Pathophysiology of cancer and the entropy concept, Model-Based Reasoning in Science and..., Metze [/bib_ref] [bib_ref] Spontaneous apoptosis in chronic lymphocytic leukemia is not an independent prognostic factor..., Metze [/bib_ref] [bib_ref] The fractal dimension of nuclear chromatin as a prognostic factor in acute..., Adam [/bib_ref] [bib_ref] Chromatin phenotype karyometry can predict recurrence in papillary urothelial neoplasms of low..., Montironi [/bib_ref] [bib_ref] Parameters derived from the fast Fourier transform are predicitive for the recurrence..., Metze [/bib_ref] [bib_ref] Prognostic classification of early ovarian cancer based on very low dimensionality adaptive..., Nielsen [/bib_ref] [bib_ref] Fractal dimension is an independent prognostic factor for survival in melanoma, Bedin [/bib_ref] [bib_ref] Proliferation in non-Hodgkin's lymphomas and its prognostic value related to staging parameters, Lorand-Metze [/bib_ref]. Therefore, we would expect that chromatin texture features in myeloma cells might reveal characteristics related to the aggressiveness of the tumor.
Scale-invariant self-similarity is an important feature of many biological structures. It cannot be described adequately by classic Euclidean geometry, but may be estimated by the determination of the fractal dimension. There is increasing use of fractal geometry for medical signal analysis with applications to pattern recognition, texture analysis and segmentation [bib_ref] The fractal geometry of life, Losa [/bib_ref] [bib_ref] Fractal characterization of chromatin appearance for diagnosis in breast cytology, Einstein [/bib_ref] [bib_ref] Fractal and multifractal analysis: a review, Lopes [/bib_ref] [bib_ref] Shannon's entropy and fractal dimension provide an objective account of bone tissue..., Rocha [/bib_ref] [bib_ref] Remodeling of the human dermis after application of salicylate silanol, Herreros [/bib_ref] [bib_ref] Fractal analysis of vascular networks: insights from morphogenesis, Lorthois [/bib_ref] [bib_ref] Fractal dimension of chromatin texture of squamous intraepithelial lesions of cervix, Dey [/bib_ref] [bib_ref] Application of the Minkowski-Bouligand fractal dimension for the differential diagnosis of thyroid..., Ferreira [/bib_ref] [bib_ref] Vessel remodelling during tumour progression of carcinoma ex pleomorphic adenoma, Metze [/bib_ref] [bib_ref] Relationship between tumor grade and computed architectural complexity in breast cancer specimens, Tambasco [/bib_ref]. This analysis has also been applied for the study of cell nuclei, as the fractal nature of chromatin and its surrounding nucleoplasmic space has been demonstrated by different methods [bib_ref] Comprehensive mapping of long-range interactions reveals folding principles of the human genome, Lieberman-Aiden [/bib_ref] [bib_ref] Molecular crowding affects diffusion and binding of nuclear proteins in heterochromatin and..., Bancaud [/bib_ref] [bib_ref] Fractal dimension of chromatin and cancer prognosis, Metze [/bib_ref]. Moreover, the fractal characteristics of nuclear chromatin measured in cytological or histological preparations have shown to be of prognostic importance in several neoplasias [bib_ref] The fractal dimension of nuclear chromatin as a prognostic factor in acute..., Adam [/bib_ref] [bib_ref] Prognostic classification of early ovarian cancer based on very low dimensionality adaptive..., Nielsen [/bib_ref] [bib_ref] Fractal dimension is an independent prognostic factor for survival in melanoma, Bedin [/bib_ref] [bib_ref] Fractal dimension of chromatin and cancer prognosis, Metze [/bib_ref] [bib_ref] Nuclear fractal dimension as a prognostic factor in oral squamous cell carcinoma, Goutzanis [/bib_ref] [bib_ref] Lymphoma and leukemia cells possess fractal dimensions that correlate with their biological..., Mashiah [/bib_ref] [bib_ref] Fractal dimension as a prognostic factor for laryngeal carcinoma, Delides [/bib_ref] [bib_ref] Fractal Analysis of Monolayer Cell Nuclei from Two Different Prognostic Classes of..., Nielsen [/bib_ref] [bib_ref] Toward objective prognostic grading of prostatic carcinoma using image analysis, Irinopoulou [/bib_ref].
Therefore we investigated whether the fractal dimension of nuclear chromatin measured in routinely stained cytological smears of myeloma patients has a relation to the survival of the patients.
# Methods
## Study subjects
We analyzed retrospectively consecutive newly diagnosed patients with multiple myeloma treated at our Institution, where clinical data, diagnostic bone marrow (BM) smears, and cytogenetic analyses were available. The study was approved by the Ethics Committee of the Faculty of Medical Sciences, University of Campinas (process 365/2002). Patients gave written informed consent for participation in the study.
MM was diagnosed using the criteria of the International Myeloma Working Group, based on serum protein electrophoresis, presence of proteinuria, bone marrow cytology and bone radiograms. The data about peripheral blood (PB) counts, bone marrow cytology, bone radiograms, serum biochemistry (calcium, albumin, creatinine and beta-2-microglobulin) and cytogenetics were recorded. Only patients that fulfilled the criteria for chemotherapy were included. Those with asymptomatic (smoldering) myeloma, immunoglobulin IgM-related disorders or with primary amyloidosis were excluded.
Staging was performed according to the ISS [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref]. Moreover, using the criteria of Greipp et al [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref] , patients were classified as belonging to the very poor prognosis group, if one or more of the following criteria were fulfilled: serum ß2-microglobulin .10 mg/ L, serum creatinine .4 mg/dL, serum albumin ,2.5 g/dL, and platelet count ,130610 9 /l. Cytogenetics had been performed in the diagnostic BM according to the method of Brigadeau et al [bib_ref] Miniaturized method for the karyotype analysis of bone marrow or blood samples..., Brigaudeau [/bib_ref]. In brief, BM nucleated cells were obtained by Ficoll-Hypaque centrifugation of the aspirated material. They were cultured for 72 hours in RPMI1640 supplemented with fetal calf serum. Cultures were performed both without stimulation, and with stimulation by IL-6 and GM-CSF. G-banding was analyzed, chromosome number and specific abnormalities were recorded.
Furthermore we looked for the presence of unfavorable cytogenetic alterations, such as hypodiploidy, deletion of chromosome , t(14:16), or 17p-, and counted for each patient the number of cytogenetic abnormalities.
The patients had been treated according to the Brazilian Multiple Myeloma Study Group [bib_ref] Thalidomide + Dexamethasone as Maintenance after Single Autologous Stem Cell Transplantation Improves..., Maiolino [/bib_ref] which consisted of an induction chemotherapy with VAD (vincristine 0.4 mg IV day 1, doxorubicin 9 mg/m 2 IV day 1 and oral dexamethasone 40 mg daily), stem cell mobilization with cyclophosphamide (4 g/m 2 IV), autologous bone marrow transplantation, and then randomized to test the utility for post-transplantation maintenance treatment using thalidomide.
Survival was measured from the diagnosis to time of death or last follow-up. Patients that underwent autologous bone marrow transplantation were censored at the day of transplantation.
## Morphometry
May-Grünwald-Giemsa stained bone marrow slides made at diagnosis were retrieved from the files. Neoplastic plasma cells were acquired in 24-bit color bitmap format with a Leica DC 500 (tm) digital camera (resolution of 12 megapixels, oil immersion objective 6100).
Randomly chosen, non-overlapping tumor nuclei were captured, interactively segmented and then converted to 8 bit gray scale with levels of luminance ranging between 0 and 255 [fig_ref] Figure 1: Segmented nucleus [/fig_ref]. For each nucleus, classical morphometric parameters such as nuclear area and the circular form factor (the ratio between perimeter of a circle with the same area as the nucleus and the actual nuclear perimeter) were calculated. In addition, the fractal dimension of each nucleus was determined.
The fractal dimension of an object is, in many studies, measured after binarization of the image. Since we were dealing with 256 gray levels, binarization would be arbitrary and reduce the information content [bib_ref] The fractal dimension of nuclear chromatin as a prognostic factor in acute..., Adam [/bib_ref] [bib_ref] Pitfalls in prognostic factor studies, Metze [/bib_ref] [bib_ref] Dichotomization of continuous data-a pitfall in prognostic factor studies, Metze [/bib_ref]. Therefore we decided not to binarize, but to apply the Minkowski-Bouligand method extended to pseudo-3D images [bib_ref] The fractal dimension of nuclear chromatin as a prognostic factor in acute..., Adam [/bib_ref] [bib_ref] Evaluating the fractal dimension of profiles, Dubuc [/bib_ref] [bib_ref] How to calculate the goodness-of-fit of a fractal dimension, Albregtsen [/bib_ref] [bib_ref] Goodness-of-fit of the fractal dimension as a prognostic factor, Metze [/bib_ref]. We created landscape-like pseudo-3D images, where the x and y coordinates represented the position of the pixel, and the z coordinate, its grey level [fig_ref] Figure 1: Segmented nucleus [/fig_ref]. The fractal dimension (FD) of the surface of the pseudo-3-D images was calculated using a software developed in house by our group. In brief, we calculated the fractal area which is estimated by the volume/2e with e being the radius, varying between 1 and 30 pixels, of the non-planar structuring element in form of a ball. Then, we calculated the linear regression in a log-log plot of the fractal area versus e, with each plot containing 30 points [fig_ref] Figure 1: Segmented nucleus [/fig_ref].We rotated the point distribution so that the slope was at 45 degrees. Then, the coefficient of the regression between the real and the estimated values was calculated [bib_ref] How to calculate the goodness-of-fit of a fractal dimension, Albregtsen [/bib_ref] [bib_ref] Goodness-of-fit of the fractal dimension as a prognostic factor, Metze [/bib_ref]. This R 2 value characterizes the goodness-of-fit of the regression line and is therefore an estimate of the ''quality of fractality''. An ideal fractal has a R 2 = 1.0. The real fractals have R 2 values ,1.0 [bib_ref] Goodness-of-fit of the fractal dimension as a prognostic factor, Metze [/bib_ref].
In a preliminary study we have found that the cumulative histograms for FD distribution in individual cases stabilized after examination of 25-35 cells [fig_ref] Figure 2: Mean fractal dimension as a function of the number of nuclei examined... [/fig_ref]. Therefore, we decided to acquire at least 40 cells per smear.
Finally, the prognostic relevance of all these parameters was analyzed in univariate and multivariate Cox regressions (p = 0.05 for input and p = 0.1 for output, backward conditional step-wise selection). The stability of the Cox model was tested by bootstrap resampling. This is a useful procedure to test the internal stability of a model proposed. It consists of creating new data sets of equal size by random sampling of the original data with replacement. In an individual new bootstrap sample, a patient may be represented once, multiple times or not at all. A new Cox regression (with the same conditions as in the original data set) was then calculated for each of these new data sets in order to obtain the bootstrap parameter estimates. This procedure is very useful in order to point out the most important variables [bib_ref] Nucleus organizer regions (AgNORs) and total tumor mass are independent prognostic parameters..., Metze [/bib_ref] [bib_ref] Density of dendritic cells around basal cell carcinomas is related to tumor..., Rybka [/bib_ref]. SPSS 8.0 and WinStat software were used for calculations.
In order to facilitate the comparison with previous studies that used plasma cell morphology, we randomly selected 107 nuclear images from our whole data pool and classified them in four categories [fig_ref] Figure 3: Segmented images of myeloma cell nuclei, classified by their grade of maturity... [/fig_ref] A: Nuclei with a mature chromatin structure; B: immature nuclei with a less organized chromatin; C: Nuclei with a loose chromatin and a large nucleolus; and D: Nuclei with irregular and disorganized chromatin with many dark and light spots. Then we compared the FDs and R 2 values between the four groups by analysis of variance followed by the least square difference post-hoc test. Differences were considered significant when p,0.05.
# Results
Sixty seven patients entered the study: 39 men and 28 women.
The clinical data are shown on table 1. Median age was 56 years. According to ISS there was a predominance of advanced cases. Additional features of a very bad prognosis were found in 46% of the patients. Among them, a b2-microglobulin value .10 mg/L was found in 37%, platelets were ,130610 9 /L in 21%, creatinine was .4 mg/dL in 15% and albumin was ,2. [fig_ref] Figure 5: Overall survival categorized by the presence or absence of additional risk factors... [/fig_ref].
Age, serum calcium concentration, hemoglobin value, PB platelet count, BM percentage of myeloma cells, nuclear area and form factor had no prognostic relevance.
The morphometric features studied are shown in table 2. The fractal dimension (FD) was an adverse prognostic feature (B = 12.33; p = 0.0287). Only in order to illustrate the importance of FD as a prognostic variable, we dichotomized the continuous variable FD according to a cluster analysis (Ward algorithm), which suggested as cut-point at FD = 2.13 and then created a Kaplan-Meier-plot [fig_ref] Figure 6: Survival of the patients grouped by FD [/fig_ref]. The goodness-of -fit (R 2 ) was a favorable prognostic factor in the univariate Cox regression when stratified for ISS (B = 2592.6; p = 0.10) [fig_ref] Table 3: Factors influencing the survival of the patients, analyzed by the univariate Cox... [/fig_ref].
Cytogenetic analysis revealed hyperdiploidy in 12 cases, alterations of chromosome 13 in 2 patients, 2 cases with translocations involving chromosome 14 (in one case together with 217) and 1 patient with hypodiploidy. The number of cytogenetic abnormalities was an adverse prognostic factor in the univariate Cox regression (B = 0.8833; p = 0.0387).
The nuclear area, form factor, FD and goodness-of-fit did not correlate with ISS stage or laboratory parameters, such as serum calcium concentration, b2-microglobulin, albumin, creatinine, hemoglobin value, PB platelet count and BM percentage of myeloma cells (Spearman rank order test). We could not detect any statistical relation between presence or number of chromosomal abnormalities and FD (Spearman rank order correlation).
Looking at the morphology of the neoplastic plasma cell nuclei [fig_ref] Figure 3: Segmented images of myeloma cell nuclei, classified by their grade of maturity... [/fig_ref] , mature plasma cells showed significantly the lowest fractal dimension (FD = 2.08) whereas the FDs of all other cell types were significantly higher (FD immature = 2.15; FD blasts = 2.18 and FD irregular nuclei = 2.16). There was, however, no statistically significant difference between these three cell types. Goodness-offit was significantly lowest for cells with irregular nuclei (R 2 = 0.
# Discussion
The median age of our patients at diagnosis was 56 years, which is considerably lower than that reported in studies on MM from North America, Europe or Asia [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref] [bib_ref] Trends in the incidence and survival of multiple myeloma in South East..., Renshaw [/bib_ref]. A younger age of the patients at diagnosis, when compared with other countries, has also been reported for other hematological neoplasias in Brazil, such as myelodysplastic syndromes and myeloid leukemias [bib_ref] Confirmation of the utility of the International Staging System and identification of..., Hungria [/bib_ref] [bib_ref] Acute leukemias in Piauí: comparison with features observed in other regions of..., Rego [/bib_ref] [bib_ref] Outcome of patients with chronic myeloid leukemia with T315l BCR-ABL mutation, Pagnano [/bib_ref] [bib_ref] Factors influencing survival in myelodysplastic sybdromes in a Brazilian population: Comparison of..., Lorand-Metze [/bib_ref] [bib_ref] Phenotypic quantitative features of patients with acute myeloid leukemia, Pereira [/bib_ref] [bib_ref] The Brazilian pediatric myelodysplastic cooperative group strategies: are they relevant to improve..., Lopes [/bib_ref] [bib_ref] Multiple myeloma in elderly patients: prognostic factors and outcome, Anagnostopoulos [/bib_ref]. Besides that, a recent epidemiological survey of patients with multiple myeloma in our country showed a similar mean age and also a high frequency of advanced ISS [bib_ref] Confirmation of the utility of the International Staging System and identification of..., Hungria [/bib_ref]. No final conclusive explanation has been found for this phenomenon. But one might speculate that this could be due to a high exposure to carcinogenic agents in Brazil. It has been shown that herbicides and pesticides are risk factors for the genesis of multiple myeloma [bib_ref] Cancer Incidence among Pesticide Applicators Exposed to Alachlor in the Agricultural Health..., Lee [/bib_ref]. But not only farm workers with direct contact to these toxins are at risk. The general population may be frequently exposed to residual toxins in food [bib_ref] Pesticides poisoning in Brazil: the official notification system and challenges to conducting..., Faria [/bib_ref] [bib_ref] Pesticide Residues in some Commodities: Dietary Risk for Children, Gebara [/bib_ref].
Our patients showed a high frequency of advanced disease according to the ISS and additional poor risk factors. This resulted in a median survival of 24.9 months for all patients, much shorter than that reported by other studies [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref] [bib_ref] Confirmation of the utility of the International Staging System and identification of..., Hungria [/bib_ref]. There are several possible explanations for this phenomenon. Delayed diagnosis in patients who depend on a public health system, as in our case, is not rare. In addition, unawareness of symptoms or indolence might be common in patients from lower socio-economic classes. And finally, a more aggressive character of MM in our country cannot be totally excluded. The patients with additional poor prognostic factors had a median survival of about 12.1 months, which is close to that reported by Greipp at al [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref]. However, only 5% of the total population reported in that study had these criteria for additional high risk.
The International Staging System (ISS), based only on two laboratory variables, the b2-microglobulin and albumin serum concentrations, has replaced the Durie-Salmon staging system in 2005 and is currently considered standard for staging of myeloma [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref] [bib_ref] Multiple myeloma: new staging systems for diagnosis, prognosis and response evaluation, Rajkumar [/bib_ref] [bib_ref] Multiple myeloma in elderly patients: prognostic factors and outcome, Anagnostopoulos [/bib_ref]. It was further validated in patients in North America, Europe, and Asia; in patients below and above 65 years of age; in patients with standard therapy or autologous bone marrow transplantation; and in comparison with the Durie/Salmon staging system. The prognostic importance of this staging system was confirmed in our study.
Regarding the great variability of myeloma pathophysiology, there are additional prognostic factors, such as age, hemoglobin concentration, serum creatinine and calcium levels. But, this could not be confirmed in our investigation. Since the number of patients was relatively small, the test power was limited and we may have missed to demonstrate them as significant risk factors.
Special attention was drawn to prognostic factors, which could improve a risk-stratification model able to define high-risk patients who can benefit from novel therapeutic strategies.
Plasma cell cytogenetic abnormalities and labeling index have also shown to be of prognostic value, but, these techniques require fresh unfixed material, are sophisticated and expensive, and can be performed only in few laboratories. Therefore it would be interesting to look for additional prognostic factors, which are not cost-expensive, can be examined retrospectively and be assessed independently of a specialized laboratory.
Cell morphology, evaluated subjectively by a trained observer, has also been considered a prognostic variable in multiple myelomas [bib_ref] Mature plasma cells as indicator of better prognosis in multiple myeloma. New..., Goasguen [/bib_ref] but the drawback of this method is the considerable inter-observer variability. Goasguen et al. developed a protocol for morphologic analysis of myeloma cells based on more objective morphologic criteria in routinely stained slides [bib_ref] Mature plasma cells as indicator of better prognosis in multiple myeloma. New..., Goasguen [/bib_ref]. These criteria included the presence of a nucleolus, blast-like chromatin and a nuclear-cytoplasmatic ratio .0.6, thus creating 8 possible subtypes. This procedure was able to identify an intermediate prognostic subgroup of patients, but the method was still dependent on a trained human observer [bib_ref] Mature plasma cells as indicator of better prognosis in multiple myeloma. New..., Goasguen [/bib_ref]. Leleu et al [bib_ref] Irregular nuclear shape of bone marrow plasma cells defines a multiple myeloma..., Leleu [/bib_ref] described in detail the nuclear shape changes in myeloma cells and created the variable ''percentage of plasma cells with irregular nuclear shape''. This variable was a prognostic factor in the univariate analysis and significantly associated with other prognostic parameters such as Ki67 labeling index, hemoglobin values and hypodiploidy, but not with beta-2-microglobulin. In both studies, however, morphologic variables were not independent risk factors in multivariate regressions.
Computerized analysis of microscopic images overcomes the necessity of morphologic expertise and expert opinion and has shown to be an objective and reproducible method for diagnostic and prognostic purposes [bib_ref] Toward objective prognostic grading of prostatic carcinoma using image analysis, Irinopoulou [/bib_ref] [bib_ref] The concept of structural entropy in tissue-based diagnosis, Kayser [/bib_ref] [bib_ref] Automatic measurement of epithelium differentiation and classification of cervical intraneoplasia by computerized..., Jondet [/bib_ref] [bib_ref] Texture-and object-related automated information analysis in histological still images of various organs, Kayser [/bib_ref]. Image analysis is able to detect subtle morphologic changes which cannot be recognized even by a trained observer [bib_ref] Application of the Minkowski-Bouligand fractal dimension for the differential diagnosis of thyroid..., Ferreira [/bib_ref] [bib_ref] Automatic measurement of epithelium differentiation and classification of cervical intraneoplasia by computerized..., Jondet [/bib_ref] [bib_ref] Texture-and object-related automated information analysis in histological still images of various organs, Kayser [/bib_ref] [bib_ref] Steroid Hormones Modify Nuclear Heterochromatin Structure and Plasma Membrane Enzyme of MCF-7..., Losa [/bib_ref]. Among these techniques, the examination of the fractal characteristics of nuclear chromatin has shown to be of increasing importance [bib_ref] Nuclear patterns of human breast cancer cells during apoptosis: characterization by fractal..., Losa [/bib_ref] [bib_ref] Using Fractal Signature Vectors and Lacunarity Class Distance Matrices to Extract New..., Nielsen [/bib_ref] [bib_ref] The influence of staining characteristics on nuclear texture features in cytology, Metze [/bib_ref] [bib_ref] Intrinsic problems of the nuclear shape factor analysis, Metze [/bib_ref] [bib_ref] Robust variables in texture analysis, Metze [/bib_ref] [bib_ref] Stochastic models for the Earth's relief, the shape and the fractal dimension..., Mandelbrot [/bib_ref].
The use of the fractal concept for image analysis has several advantages. The fractal dimension has shown to be robust against the segmentation process [bib_ref] Robust variables in texture analysis, Metze [/bib_ref]. In routine cytological preparations fractal derived variables are much less dependent on staining variations than variables derived from the grey-level cooccurrence matrix [bib_ref] The influence of staining characteristics on nuclear texture features in cytology, Metze [/bib_ref]. Form factors, which are classically used for the quantitative description of irregular outlines, depend highly on the magnification scale, whereas fractals are scale independent [bib_ref] Intrinsic problems of the nuclear shape factor analysis, Metze [/bib_ref].
The fractal dimension represents a statistical description but, moreover, is also intimately related to the theoretical concepts of complexity and morphogenesis. Therefore it provides a deeper insight into the understanding of the biology of normal tissues and neoplasias [bib_ref] The fractal geometry of life, Losa [/bib_ref] [bib_ref] Fractal dimension of chromatin and cancer prognosis, Metze [/bib_ref]. The fractal concept, developed by Mandelbrot [bib_ref] Stochastic models for the Earth's relief, the shape and the fractal dimension..., Mandelbrot [/bib_ref] in the sixties and seventies of the last century got popularity due to the famous images created by computer programs based on fractal geometry. It is however a ubiquitous theoretical framework for many processes or objects in our known universe. Fractality implies scale-independent self-similarity or self-affinity. The fractal concept may be applied to every kind of science. Fractal characteristics can be found when measurement values of a certain variable are scale independent so that in a log-log plot the measurement points can be well approximated by a regression line. Therefore the fractal property is always related to a measurement variable. An ''object'' or a ''process'' can reveal simultaneously fractal characteristics of many different variables or features. Introducing the fractal concept in biology and medicine has improved our understanding of many physiological processes, such as allometric scaling growth, allosteric enzyme kinetics, intracellular bio-energetic dynamics, metabolic rate in mammals, population genetics, modeling of drug clearance, neo-angiogenesis, tumor growth, organization of nucleotides in DNA and RNA, and cardiovascular physiology [bib_ref] The fractal geometry of life, Losa [/bib_ref] [bib_ref] Fractal and multifractal analysis: a review, Lopes [/bib_ref] [bib_ref] Fractal analysis of vascular networks: insights from morphogenesis, Lorthois [/bib_ref] [bib_ref] Complexity on acute myeloid leukemia mRNA transcript variant, Cattani [/bib_ref] [bib_ref] The Scale-Free Dynamics of, Aon [/bib_ref] [bib_ref] Fractals for physicians, Thamrin [/bib_ref] [bib_ref] Fractal geometry in the nucleus, Mcnally [/bib_ref]. Fractals are very useful to characterize properly the complexity of macroscopic and microscopic anatomy, namely to describe the design principles underlying living organisms.
The chromatin structure can also be analyzed by fractal geometry [bib_ref] The fractal dimension of nuclear chromatin as a prognostic factor in acute..., Adam [/bib_ref] [bib_ref] Fractal dimension is an independent prognostic factor for survival in melanoma, Bedin [/bib_ref] [bib_ref] The fractal geometry of life, Losa [/bib_ref] [bib_ref] Fractal characterization of chromatin appearance for diagnosis in breast cytology, Einstein [/bib_ref] [bib_ref] Fractal dimension of chromatin texture of squamous intraepithelial lesions of cervix, Dey [/bib_ref] [bib_ref] Application of the Minkowski-Bouligand fractal dimension for the differential diagnosis of thyroid..., Ferreira [/bib_ref] [bib_ref] Robust variables in texture analysis, Metze [/bib_ref] [bib_ref] Stochastic models for the Earth's relief, the shape and the fractal dimension..., Mandelbrot [/bib_ref] [bib_ref] Complexity on acute myeloid leukemia mRNA transcript variant, Cattani [/bib_ref] [bib_ref] The Scale-Free Dynamics of, Aon [/bib_ref] [bib_ref] Fractals for physicians, Thamrin [/bib_ref] [bib_ref] Fractal geometry in the nucleus, Mcnally [/bib_ref].
Fractal structures can be created by iterations. This happens for instance during condensation of a polymer such as DNA when it is repeatedly subjected to the self-similar process of crumpling. The result is a folded polymer with fractal properties. In this way, a long polymer can be packed in a small volume without entanglements. This facilitates unravelling when necessary, as for DNA transcription, and is therefore advantageous for cell physiology [bib_ref] Fractal geometry in the nucleus, Mcnally [/bib_ref]. Recent studies using different methods of analysis showed that DNA, nuclear chromatin and the surrounding nucleoplasmic space have a fractal organization [bib_ref] Comprehensive mapping of long-range interactions reveals folding principles of the human genome, Lieberman-Aiden [/bib_ref] [bib_ref] Molecular crowding affects diffusion and binding of nuclear proteins in heterochromatin and..., Bancaud [/bib_ref] [bib_ref] Fractal dimension of chromatin and cancer prognosis, Metze [/bib_ref] [bib_ref] Fractal geometry in the nucleus, Mcnally [/bib_ref] ,
Our study revealed that the staining pattern of nuclear chromatin of myeloma cells, (May-Grünwald -Giemsa method) has also fractal characteristics. An important challenge is to explain, why patients with a worse prognosis, revealed an increased FD of the chromatin staining pattern in May-Grünwald -Giemsa stained smears. Changes of the nuclear architecture, observed in histological or cytological preparations, reflect genomic and non-genomic alterations.
Multiple genetic aberrations have been described during the pathogenesis of multiple myelomas [bib_ref] Multiple myeloma, Raab [/bib_ref]. Secondary translocations and gene mutations have been implicated in disease progression, such as complex abnormalities of MYC, activation of N-RAS, K-RAS, and FGFR3 mutations, mutations or deletions of TP53, RB1, and PTEN and inactivation of cyclin-dependent kinase inhibitors [bib_ref] Epigenetic Silencing of theTetraspanin CD9 during Disease Progression in Multiple Myeloma Cells..., De Bruyne [/bib_ref] [bib_ref] Differential repetitive DNA methylation in multiple myeloma molecular subgroups, Bollati [/bib_ref].
Furthermore, epigenetic changes causing altered gene and protein expression play a major role in the pathogenesis of multiple myelomas. Hypermethylation of the genes p15, p16, DAP-kinase, BAD, BAK, BAX, BIK, SOCS-1, and E-Cadherin, has been reported [bib_ref] Epigenetic Silencing of theTetraspanin CD9 during Disease Progression in Multiple Myeloma Cells..., De Bruyne [/bib_ref] [bib_ref] Differential repetitive DNA methylation in multiple myeloma molecular subgroups, Bollati [/bib_ref].
Hypermethylation of CpG islands within gene promoter regions accompanied by deacetylation of histone proteins provokes transcriptional silencing. In parallel, global hypomethylation of repetitive elements occurs in association with tumor progression and increase of chromosomal instability [bib_ref] Epigenetic Silencing of theTetraspanin CD9 during Disease Progression in Multiple Myeloma Cells..., De Bruyne [/bib_ref] [bib_ref] Differential repetitive DNA methylation in multiple myeloma molecular subgroups, Bollati [/bib_ref]. Therefore, in advanced cases, which are genetically unstable and therefore more aggressive, we expect a more complex chromatin rearrangement with global hypomethylaton and an increased number of hypermethylated CPG islands.
The routinely MGG stained cytologic smears permit to estimate the topographic localization of methylated regions in the nucleus. Co-localization analysis has shown that the deeply Giemsa-stained compacted heterochromatin domains have the same geographic distribution as the methyl-rich regions within each nucleus [bib_ref] Reduced levels of poly(ADP-ribosyl)ation result in chromatin compaction and hypermethylation as shown..., De Capoa [/bib_ref]. Therefore, aggressive myelomas should have a more complex chromatin structure with an increased number of darker and lighter areas, leading to a higher fractal dimension of the nuclear chromatin, as found in our study.
Indeed, the FD of MGG-stained nuclear chromatin showed to be an independent adverse prognostic factor for the overall survival of our patients. In a similar way, previous studies have demonstrated an association between a higher FD value of the nuclear chromatin and a worse outcome of patients with other neoplasias, such as malignant melanomas, squamous cell carcinomas of the oral cavity and larynx [bib_ref] Fractal dimension is an independent prognostic factor for survival in melanoma, Bedin [/bib_ref] [bib_ref] Fractal dimension of chromatin and cancer prognosis, Metze [/bib_ref] [bib_ref] Nuclear fractal dimension as a prognostic factor in oral squamous cell carcinoma, Goutzanis [/bib_ref] [bib_ref] Lymphoma and leukemia cells possess fractal dimensions that correlate with their biological..., Mashiah [/bib_ref] [bib_ref] Fractal dimension as a prognostic factor for laryngeal carcinoma, Delides [/bib_ref].
This implies that the complexity of chromatin distribution contains important prognostic information which is independent of clinical variables, ISS stage, or relevant cytogenetic aberrations. In the present study, the frequency of cytogenetic abnormalities was low when compared to other reports [bib_ref] International staging system for multiple myeloma, Greipp [/bib_ref] [bib_ref] Chromosomal translocations t(4;14), t(11;14) and proliferation rate stratify patients with mature plasma..., Tinguely [/bib_ref]. Technical problems, however, could have precluded the finding of some abnormalities.
The goodness-of-fit of the fractal dimension was also of independent prognostic relevance, but as a favorable factor. Thus a chromatin architecture closer to the "ideal fractal" was associated with a prolonged survival of the patient, as it has been shown previously for blasts of B precursor acute lymphatic leukemia [bib_ref] The fractal dimension of nuclear chromatin as a prognostic factor in acute..., Adam [/bib_ref].
Explanations for this finding are speculative at the moment. But it might be possible that the large number of genetic and epigenetic modifications in aggressive tumors could disturb the process of auto-organization of the nucleus to such an extent that the scale-free auto-similarity of chromatin structures is not as perfect as in less aggressive cases.
Our study revealed that both the fractal dimension and its goodness-of-fit permit to quantify DNA remodeling and methylation status in MGG-stained bone marrow smears of myeloma patients and therefore may be new and biologically relevant prognostic factors for this disease. A more detailed scientific validation of the texture analyses is necessary, of course. Data from genome-wide methylation or chromatin analyses should be compared with fractal data derived from digitalized images of routinely stained tumor smears or sections, in order to define better the equivalence between changes of the image texture and genome-wide biochemical chromatin changes.
Since there is a vast heterogeneity of molecular profiles among myeloma patients, detailed individual genomic evaluations for targeted therapies seem not to be helpful at the moment, as has been emphasized recently [bib_ref] Molecular Heterogeneity of Multiple Myeloma: Pathogenesis, Prognosis, and Therapeutic Implications, Avet-Loiseau [/bib_ref].
In this situation, a global evaluation of the nucleus, as presented in this investigation, could be interesting. This technique is simple, reproducible and unexpensive and may be applied to routine slides from the files, thus permitting retrospective studies without any additional costs. Therefore, we think it could be useful in daily routine practice in future. But since our study was based on a relatively small number of patients, it should, of course, be followed by confirmatory investigations based on more and new patients in different centers [bib_ref] The importance of the mitotic index as a prognostic factor for survival..., Elston [/bib_ref].
[fig] Figure 1: Segmented nucleus (left) of a myeloma cell and its pseudo 3D transformation (middle). On the right side, the log-logplot for the determination of the fractal dimension (FD), which is calculated from the slope of the ideal regression line (black) obtained by curve fitting. X-axis shows the logarithms of the inverse values of the size of the structuring element and y axis the logarithmic values of the fractal areas. (compare with the main text). R 2 represents the goodnessof-fit of the real values (red), when compared with ideal regression line (black) and can be interpreted as a measure of ''fractal quality''. doi:10.1371/journal.pone.0020706.g001 [/fig]
[fig] 5: g/dL in 19% of the patients. The mean follow-up was 25.4 months (1-227). At the end of the observation period 36 patients had died from multiple myeloma, 24 of them before having completed the VAD regimen. ISS (Fig. 4) was a significant adverse prognostic factor in the univariate Cox regression (stage I: B = 22.30; stage II B = 21.50; stage III: B = 0.0; p = 0.002). Patients belonging to the very poor prognostic group had a significant shorter survival (median 12.1 [/fig]
[fig] Figure 2: Mean fractal dimension as a function of the number of nuclei examined in an individual case. The mean value stabilized after examination of 30-40 cells. doi:10.1371/journal.pone.0020706.g002 [/fig]
[fig] Figure 3: Segmented images of myeloma cell nuclei, classified by their grade of maturity and chromatin structure. A: Nucleus with a more mature chromatin structure; its FD was 2.074935 and R 2 = 0.999330. B: a more immature nucleus with a less organized chromatin structure presenting FD = 2.223 and R 2 = 0.99896. C: Nucleus with a blastic feature: FD = 2.1759 and R 2 = 0.9961. D: Nucleus with irregular and disorganized chromatin: FD = 2.2629 and R 2 = 0.99931. doi:10.1371/journal.pone.0020706.g003 [/fig]
[fig] Figure 4: Overall survival of the myeloma patients categorized by ISS stage. Kaplan-Meier plot. Log rank-test: p = 0.0005. doi:10.1371/journal.pone.0020706.g004 [/fig]
[fig] Figure 6: Survival of the patients grouped by FD. Kaplan-Meier plot. Log rank-test: p = 0.05. doi:10.1371/journal.pone.0020706.g006 [/fig]
[table] Table 1: Clinical variables of the patients. [/table]
[table] Table 2: Morphometric variables of the nuclei of the myeloma cells. [/table]
[table] Table 3: Factors influencing the survival of the patients, analyzed by the univariate Cox regression. [/table]
|
Blueberry as an Attractive Functional Fruit to Prevent (Pre)Diabetes Progression
Citation: Nunes, S.; Vieira, P.; Gomes, P.; Viana, S.D.; Reis, F. Blueberry as an Attractive Functional Fruit to Prevent (Pre)Diabetes Progression.
# Introduction
The type 2 diabetes mellitus (T2DM) epidemic is a major public health concern. According to the International Diabetes Federation (IDF) estimates, 463 millions of people are living with diabetes globally, and it is projected to rise to 700 million by 2045. Paralleling this daunting scenario, prediabetes (also referred as intermediate hyperglycemia by the World Health Organization (WHO)is also an increasingly common condition, characterized by blood glucose levels above the normal range but below diabetes diagnostic thresholds, and typically defined as a state of impaired fasting glucose (IFG) and/or impaired glucose tolerance (IGT). While the IDF has no data about the prevalence of IFG, 373 million people worldwide are estimated to have IGT, a number that is predicted to increase to 548 million by 2045.
Although the prevalence of prediabetes varies according to the criteria used for diagnosis (WHO vs. American Diabetes Association, ADA) and the study population, it is consensually recognized as a risk state for T2DM development. In fact, both insulin resistance and impaired β-cell function, which are disease hallmarks and early pathophysiological mechanisms, are responsible for the increased risk of T2DM development in patients with prediabetes. Evidence points to a yearly conversion rate of 5-10%, but also to cardiovascular disease (CVD) and even cancer. Multiple genetic and environmental risk factors contribute to the progression of the disease, namely the age of the population, its genetic background, a sedentary lifestyle and poor (hypercaloric) nutritional habits, collectively leading to obesity, hepatic steatosis and dyslipidemia, which are important predisposing factors for the progression of T2DM and CVD.
Along with peripheral insulin resistance and defects in insulin secretion, other unbalanced metabolic processes, including glucotoxicity and lipotoxicity, oxidative stress and low-grade inflammation are already evolving in the state of prediabetes. In addition, recent evidence suggests that changes in gut microbiota (GM) community may also play a role in T2DM pathophysiology and impaired composition and diversity of intestinal bacterial flora (referred as dysbiosis) may actually begin at an early stage.
Lifestyle therapy, including regular physical activity and diabetes-specific nutrition therapy, is the cornerstone for the early management of prediabetes due to their accessibility, cost-effectiveness as well as the limited pharmacological options with favorable safety profiles namely in the earlier stages of the disease. Accordingly, prospective and meta-analyses studies revealed an inverse correlation between fruit and vegetable intake and the increased risk and incidence of T2DM. Moreover, the potential beneficial roles of phytochemicals (e.g., polyphenols, prebiotic fibers) and macronutrients in plantbased diets for the prevention of metabolic syndrome and T2DM are emerging. Among various bioactive components, epidemiological studies and randomized clinical trials clearly demonstrate that diets with high-content of polyphenolic compounds, high antioxidant capacity and prebiotic/probiotic/symbiotic activities are inversely associated with the risk of insulin resistance and (pre)diabetes progression. Blueberries (BB), due to their enriched content in the mentioned phytochemicals, have enticed keen interest as a feasible dietary intervention for the prevention and management of (pre)diabetes. Notably, both observational and clinical studies as well as cumulative preclinical data strongly underscore BB as an attractive complementary strategy to halt the progression from prediabetes to overt T2DM. Besides their well-recognized hypoglycemic and insulin-sensitizing effects, BB-enriched diets further improve GM homeostasisand lessen hepatic dysmetabolism, two key components of (pre)diabetes evolution.
In this work, we review current evidence linking the benefits of consuming BB with hepatic-and gut-derived pathophysiological traits ensuing as soon as prediabetes. Moreover, we provide an updated analysis of the main cellular and molecular mechanisms by which BB phytochemicals are able to exert protective effects.
Studies search was performed in PubMed and ClinicalTrials.gov databases using a strategy based on keywords. Most consistent, relevant and well-designed studies (both original and literature reviews) published between 2000-2021 were included. Collectively, this work scrutinizes forefront evidence in support of BB-enriched diets as an attractive option to prevent (pre)diabetes progression.
## From prediabetes to overt diabetes-emphasis on liver and gut paths
The transition from a normal state to prediabetes, and then to diabetes, is the consequence of a series of pathophysiological changes, that makes the individual gradually more susceptible to the subsequent disruption of glucose homeostasis, which is mainly governed by insulin sensitivity of peripheral tissues and pancreatic β-cell function. Although there has been a significant debate in the past on whether insulin resistance or β-cell dysfunction is the primary driver in the pathogenesis of prediabetes as well as on their relative contributions to disease development, the current view is that both conditions dynamically influence each other and presumably synergistically exacerbate diabetes. Increased peripheral insulin resistance and altered β-cell function leads to major biochemical alterations in metabolic tissues, including the liver, gut, skeletal muscle and adipose tissue, among other organs. Notably, oxidative stress and low-grade chronic inflammation seem to be major molecular mechanisms underlying (pre)diabetes progression, both as a cause or consequence of hyperglycemia and hyperinsulinemia.
In addition to being the main detoxifying organ of the body, the liver also plays a fundamental role in maintaining energy and metabolic homeostasis. Disturbances in hepatic glucose and lipid metabolism similarly contribute to the development of metabolic diseases, including T2DM and non-alcoholic fatty liver disease (NAFLD). Because glucose is a readily available energy source for cells, the maintenance of blood glucose levels within a relatively narrow range during periods of nutrient shortage or excess is critical for survival. The liver plays a major role in maintaining whole body glucose homeostasis by regulating several metabolic processes involved in glucose transport, oxidation, storage and production. Simultaneously, hepatic lipid metabolism encompasses fatty acid uptake, synthesis, oxidation, storage, and secretion, which requires a well-orchestrated balance between these processes, dietary intake and energy expenditure. Hepatic steatosisthe hallmark and earliest feature of NAFLD-is defined as an excessive accumulation of TGs in the cytoplasm of hepatocytes due to imbalances in the processes that maintain the normal hepatic fatty acid metabolism, and often occurs in the setting of overnutrition and obesity. Collectively, the data available support the notion that hepatic steatosis is a favorable biological milieu for the development and progression of (pre)diabetes. Several potential mechanisms could explain the association between hepatic steatosis and the onset of prediabetes, including increased levels of hepatokines and the presence of hepatic insulin resistance.
Hundreds of trillions of microorganisms inhabit the human body, the overwhelming majority bacteria from two main phyla (Bacteroidetes and Firmicutes). This complex bacterial community is present in many body fluids and surfaces but has a huge implantation in the gastrointestinal tract (GIT), in particular in the large intestine. GM can be seen, in a way, as a supra-organism that performs crucial functions for host homeostasis (co-metabolism), which include metabolic and energetic effects, protection against pathogens, maintenance of intestinal barrier integrity and education of immune system. Disturbances in the composition and/or functions of GM cause a situation referred as dysbiosis which has been associated not only with disorders of the GIT, but also with extra-intestinal diseases, including metabolic disorders such as obesity and diabetes. Accumulating evidence indicates that GM is involved in host metabolism by increased energy extraction, immune system modulation, and altered lipid and glucose metabolism, all which have been demonstrated to contribute to progression to T2DM. It has been also shown that an individual's glucose response after a meal is influenced by a combination of their GM composition and host physiology. Several studies have reported decline of butyrate-producing bacteria in prediabetes cohorts of distinct ethnicities, and to a greater extent in people with both IGT and IFG. In sum, these findings suggest causal links between microbiota dysbiosis and glucose intolerance/progression to T2DM.
## Blueberries beneficial outcomes in (pre)diabetes progression-a multi-modal functional fruit
## Bb phytochemistry, metabolism and bioavailability-an overview
Blueberry is one of the five healthy fruits recommended by the Food and Agriculture Organization from United Nations. It belongs to the genus Vaccinium of Ericaceae family which includes approximately 450 species all over the world. Some examples are the well-known Vaccinium corymbosum, Vaccinium ahei, Vaccinium poasanum, Vaccinium angustifolium and Vaccinium myrtillus, the later commonly denominated as bilberries across Europe. BB are characterized by their enriched nutritive composition, containing high contents of a wide diversity of biologically active components. BB chemical composition greatly varies upon the cultivar, variety, growing location, environmental conditions, plant nutrition, ripeness stage, harvest time as well as the storage conditions, making the content of each individual component highly flexible, with impact on the corresponding antioxidant profiles. Nonetheless, BB-derived phytochemicals display a panoply of health-related properties beyond their antioxidant profile able to interfere with chief physiological functions. Therefore, public interest in BB as a functional food has increased over the last years. Even though the aerial parts of the shrubs are used as a folk medicine for treating metabolic diseases for years, the fruits have been constantly increasing in popularity due to their pleasant taste and organoleptic acceptance as well as the richness of their bioactive compounds.
BB fruits are low in calories (0.046 kcal/g fresh fruit) and high in water, micronutrients (e.g., selenium, zinc, iron), dietary prebiotic fibers (3-3.5% of fruit weight), vitamins (C, B complex, E, and A) and sugar units [e.g., glucose (≈26.415 mg/g-46.495 mg/g fruit), fructose (≈22.682 mg/g-43.074 mg/g/fruit) and other osidic moieties such as ribose, rhamnose, arabinose and maltose] which are often associated with bioactive polyphenols (PP), known as heterosydic forms of PP. Briefly, PP are secondary metabolites of the plants that act against ultraviolet radiation as well as in plant-microbe interactions and defense response. Collectively, they encompass an aromatic ring and a minimum of two hydroxyl substituents which impact ROS scavenging ability and subsequent antioxidant effects depending on their number and position. PP comprise a large family of compounds and are classified as flavonoids (including flavonols, flavanols, flavones, isoflavones, anthocyanidins and flavanones), phenolic acids (including benzoic acids, cinnamic acids and derivatives), chalcones and coumarins (intermediate in the biosynthesis of flavonoids), and lastly, polymers, including tannins and lignin. PP are present in distinct parts of the BB fruit; for instance, while tannins and phenolic acids are mostly predominant in the seeds, anthocyanins and proanthocyanidins are present mainly in the fruit's skin and pulp, respectively. Moreover, PP occur in vegetable cells in soluble forms (mostly localized in the cell vacuoles) as well as covalently bound with cell wall macromolecules such as structural proteins, cellulose, hemicellulose, arabinoxylans and/or pectins, collectively designated as PP insoluble forms. Likewise, insoluble PP naturally present in BB fruits may become tightly linked (e.g., esther; C-C bonds) with pectin and xyloglucan, two main components of BB cell wall with prebiotic activity.
The biological properties of PP found in BB fruits greatly depend on effective concentrations reached in internal compartments that need to be maintained for an adequate period of time. Notably, the poor bioavailability of PP present in ingested food, ranging from 1-5% of the total PP intake, is inversely correlated with their biological effects, a phenomenon stated as the low bioavailability/high bioactivity paradox. Even though PP are extensively metabolized within enterocytes, colonic GM and liver by phase I/II enzymatic reactions (e.g., sulfation, methylation, glucuronidation, PP ring-fission), a wide variety of new chemical structures retaining intrinsic bioactivity along with parent compounds reach systemic circulation and are distributed to different organs and tissues, supplying efficient cellular concentrations that underpin PP efficacy.
PP physicochemical properties, their release from the food matrix during GIT digestion (bioaccessibility), cellular uptake, metabolism and transport in the circulatory system are key features that determine PP organism availability. To better understand blueberry PP metabolism and pharmacokinetics, Zhong and colleagues carried out a single blind, randomized trial and followed human plasma concentrations of anthocyanins, chlorogenic acid and their metabolites over 24 h after oral ingestion of a wild blueberry beverage. The total bioavailability of unaltered anthocyanin compounds (displaying their original C6-C3-C6 structure) and chlorogenic acid was 1.1% and 0.2%, respectively. Parent compounds and metabolites (e.g., cyanidin-, delphinidin-, petundin-glucoronide metabolites) peaked in plasma within the first 1-4 h post-ingestion (early phase response) and >5 h after ingestion (late phase response). This bi-phasic response, probably reflective of enterohepatic circulation, may imply that different BB-derived PP and metabolites interact with GIT cells and other tissues at different times post-consumption. As a matter of fact, nanostructures comprising PP-BB matrix conjugates may be protected along the transit through the GIT, a phenomenon affecting both PP absorption and cellular uptake. Likewise, in vitro gastrointestinal digestion of BB fruits underscored a high stability of total PP and anthocyanins during the gastric digestion step along with decreased contents upon intestinal breakdown (49% and 15%, respectively). This profile paralleled unchanged antioxidant activity following gastric digestion and a significant reduction (<50%) upon digestion under intestinal conditions.
Furthermore, it is generally accepted that the remaining unabsorbed PP glycosides and conjugates cycled through hepatic and ileal metabolism reach the colon and modulate fermentation towards a myriad of health benefits. Since small intestine lacks pectinases and cellulases, PP-dietary fibers conjugates are disrupted by colonic GM enzymes and PP metabolites liberation occurs. Such metabolites display chief functions in colon health, modifying gut microbial balance towards beneficial bacteria prevalence and SCFAs production. In vitro colonic digestion has also shown BB polyphenols biotransformation to syringic, cinnamic, caffeic and protocatechuic acids. Likewise, microbial metabolism of blueberry PP was found to elicit positive outcomes, namely the inhibition of HT-29 colon cancer cell line proliferation as well as the repression of colonic prostanoid production and anti-inflammatory effects. Overall, the in vivo bioactivity of blueberry polyphenols is strictly correlated with their absorption along the GIT as well colonic microbial metabolism.
## Hypoglycemic and insulin sensitizing effects of bb
A number of in vitro studies strongly suggests that BB and its bioactive phytochemicals present anti-diabetic properties. The ability of an extract from Lowbush blueberry fruit to increase pancreatic β-cell proliferation was explored in a cell culture-based bioassay, suggesting a potential capacity to restrain β-cell damage and improve insulin sensitivity. Likewise, whole BB afforded pancreatic β-cell protection and prevented β-cell apoptosis and expansion in HFD-fed C57BL/6J mice. Notably, a tight regulation of pancreatic islet area was observed as blueberry supplementation induced an increased number of small islets (comprising more β-cells with higher insulin) along with a decreased density of larger islets, probably delaying the overwhelmed burden of β-cells activity that pair obesity and diabetes progression. Similar pro-proliferative effect was observed for a BB leaf extract in pancreatic MIN6 β-cells, accompanied by improved insulin signaling. In vivo, this extract was able to decrease BW, plasma glucose, HbA1c, HOMA-IR, TGs and NEFAs levels in C57BL/6J mice fed with HFD. These effects paralleled an increased expression of pancreatic β-cell proliferation-related genes (Ngn3, MafA, Pax4, Ins1, and Ins2) and insulin signaling genes (IRS-1, IRS-2, PIK3ca, PDK1, PKCε, and GLUT-2) while FoxO1, a β-cell apoptosis-related gene, was found downregulated. Similar results were observed in male Wistar rats submitted to a combined STZ+HFD paradigm of experimental T2DM.
Furthermore, it has been reported that BB and their bioactive compounds exhibit inhibitory activity toward pancreatic α-amylase and intestinal α-glucosidase, two enzymes involved in the metabolism of starch, which could lead to a delay of carbohydrate digestion and/or decreased glucose absorption. Interestingly, in vitro studies have also highlighted that polyphenols from BB can act as strong DPP-4 inhibitors, analogous to clinically approved pharmacological options (e.g., gliptins).
Protein tyrosine phosphatase 1B (PTP1B) is another promising pharmacological target for the treatment of T2DM as it can reduce blood glucose levels by increasing insulin sensitivity. Notably, Tian et al.identified anthocyanins isolated from BB as selective inhibitors of PTP1B able to increase glucose consumption on HepG2 (human hepatoma) cells in a dose-dependent manner. Moreover, Nachar et al.performed an in vitro study with phenolic compounds from BB juice fermented with Serratia vaccinii and observed a strong antidiabetic potential in the liver and skeletal muscle, as it was able to regulate key enzymes involved in glycogen synthesis, gluconeogenesis and skeletal muscle glucose uptake. Protein tyrosine phosphatase 1B (PTP1B) is another promising pharmacological target for the treatment of T2DM as it can reduce blood glucose levels by increasing insulin sensitivity. Notably, Tian et al.identified anthocyanins isolated from BB as selective inhibitors of PTP1B able to increase glucose consumption on HepG2 (human hepatoma) cells in a dose-dependent manner. Moreover, Nachar et al.performed an in vitro study with phenolic compounds from BB juice fermented with Serratia vaccinii and observed a strong antidiabetic potential in the liver and skeletal muscle, as it was able to regulate key enzymes involved in glycogen synthesis, gluconeogenesis and skeletal muscle glucose uptake.
In vivo pre-clinical and clinical studies also hint for T2DM improvements upon BB consumption. Regarding animal models, phenolic compounds from a fermented BBblackberry beverage were able to attenuate fasting blood glucose and the development of obesity in C57BL/6J mice. Vendrame et al.also reported that wild BB consumption significantly decreased plasmatic concentrations of HbA1C, resistin, and RBP4 in the obese Zucker rat model of metabolic syndrome along with the repression of hepatic resistin and RBP4 in the abdominal adipose tissue.
Some clinical trials aimed to analyze the potential anti-diabetic effects have also been explored. In T2DM male volunteers, a single oral capsule of 0.47 g standardized bilberry extract (36% w/w anthocyanins), equivalent to 50 g of fresh bilberries, significantly decreased the area under the curve (AUC) for both glucose and insulin in 18%, with no changes on gut incretin hormones (GIP and GLP-1) nor in glucagon and amylin secreted from the pancreas. Moreover, an intervention study showed that In vivo pre-clinical and clinical studies also hint for T2DM improvements upon BB consumption. Regarding animal models, phenolic compounds from a fermented BBblackberry beverage were able to attenuate fasting blood glucose and the development of obesity in C57BL/6J mice. Vendrame et al.also reported that wild BB consumption significantly decreased plasmatic concentrations of HbA1C, resistin, and RBP4 in the obese Zucker rat model of metabolic syndrome along with the repression of hepatic resistin and RBP4 in the abdominal adipose tissue.
Some clinical trials aimed to analyze the potential anti-diabetic effects have also been explored. In T2DM male volunteers, a single oral capsule of 0.47 g standardized bilberry extract (36% w/w anthocyanins), equivalent to 50 g of fresh bilberries, significantly decreased the area under the curve (AUC) for both glucose and insulin in 18%, with no changes on gut incretin hormones (GIP and GLP-1) nor in glucagon and amylin secreted from the pancreas. Moreover, an intervention study showed that daily consumption of a BB smoothie (45 g of blueberry bioactive, equivalent to~2 cups of fresh BB per day) for 6 weeks significantly improved insulin sensitivity in thirty-two obese, insulin resistant adults, despite unchanged fasting serum glucose levels. Nevertheless, recent clinical trials reported inconsistencies on the efficacy of BB in ameliorating glucose profile in both T2DM and insulin resistant subjects. A double-blind, randomized controlled trial (RCT) was carried out to evaluate the effect of BB consumption on cardiometabolic parameters in type 2 diabetic men. Twenty-six participants consuming a beverage made from 22 g of freeze-dried BB (equivalent to 1 cup of fresh BB) daily for 8 weeks had lower HbA1c and fructosamine levels (two biomarkers of glycemic control) along with lower triglyceridaemia when compared to participants consuming the placebo. However, no significant changes in total body weight, blood pressure, lipid profile (total cholesterol, LDL and HDL-cholesterol) as well as in serum fasting glucose/insulin concentrations were detected. All of these participants were taking noninsulin diabetes medication and had a good glycemic control. In another study based on a long-term RCT, thirtyseven adults with metabolic syndrome received for 6 months a single-serve sachets with 26 g of freeze-dried BB (1 cup of BB, equivalent to 150 g of fresh BB) daily. Improved markers of vascular function and elements of lipid status (increased HDL-cholesterol and alipoprotein A-I) were detected following 1 cup of blueberries daily for 6 months. However, no benefits were observed on insulin resistance and sensitivity, measured by HOMA-IR, quantitative insulin sensitivity index (QUICKI) and HbA1c and by the gold standard 2-step, hyperinsulinemic euglyceminc clamp. This ambiguity between human intervention studies may be due to the heterogeneity of study population, the duration of the study and/or with the specific analytical technique used to assess the glycemic control. In addition, variations in sources and doses of BB-derived phytochemicals, distinct food or beverage matrix or processing may be contributing to the inconsistent findings when assessing the effects on metabolic outcomes. Notably, it is also challenging to identify the optimal dose of polyphenols needed to improve metabolic health since the amount of bioactive compound present in the aforesaid BB paradigms greatly varies between published studies.
## Antioxidant and anti-inflammatory properties of bb
BB owe their antioxidant activities mainly to their phenolic compounds including phenolic acids, condensed tannins and principally, anthocyanins. Several studies point to linear correlation between antioxidant activity and the total phenolic concentrations as well as anthocyanins in BB. Besides anthocyanins, cinnamic acid derivatives of phenolic acids (e.g., caffeic and chlorogenic acid) were found to be more active antioxidants than benzoic acids derivatives.
Chronic hyperglycemia leads to a prooxidant and inflammatory scenario. Conversely, increased oxidative stress and inflammation foster insulin resistance and impaired insulin secretion. Hence, treatments aimed to inhibit ROS overproduction and inflammatory markers are welcome to delay the onset and progression of T2DM and related complications.
Health beneficial effects of BB phytochemicals are well recognized on the basis of their potent antioxidant and anti-inflammatory properties, particularly in metabolic conditions such as T2DM. In diabetic human aortic endothelial cells, BB phytochemicals were able to restore cell surface glycosaminoglycan, attenuate endothelial inflammation, suppress monocyte binding and reduce endothelial IL-8 and vascular cell adhesion molecule 1 (VCAM-1). Additionally, BB anthocyanins extract (BAE) was able to decrease ROS generation while increasing catalase and superoxide dismutase (SOD) activities in high glucose-induced injury of human retinal capillary endothelial cells (HRCECs); furthermore, BAE was also found to influence angiogenesis by decreasing vascular endothelial growth factor (VEGF), intercellular adhesion molecule-1 (ICAM-1) and Akt pathways inhibition.
Several in vivo studies highlighted the antioxidant and anti-inflammatory effects of BB and its specific compounds on metabolic impairments. An overall improvement in the inflammatory status was reported in obese Zucker rats supplemented with wild BB powder. The consumption of 8% wild blueberry-enriched diet for 8 weeks significantly decreased plasma levels of TNF-α, IL-6 and CRP and increased adiponectin. In addition, the expression of CRP was reduced in the liver, while TNF-α, IL-6 and NF-κB were down-regulated in both liver and abdominal adipose tissue. In male Wistar rats, BB supplementation in the diet elicited the normalization of TNF-α and IL-1β levels, which were increased by 300% and 500% due to a high-fat feeding paradigm. Other authors found that BB powder afforded protection against inflammation of adipose tissue by preventing the upregulation of inflammatory genes (i.e., TNF-α and IL-10) and attenuating oxidative stress (with an increased glutathione peroxidase gene expression), collectively ameliorating obesity-induced insulin resistance. In addition, obese Zucker rats treated during 15 weeks with BB showed improved glucose tolerance and renal function, together with amelioration of oxidative balance (upholding SOD/catalase antioxidant levels while attenuating oxygen and nitrogen free radical production) and inflammatory pathways (decreased gene and protein expression of TLR4, phosphorylation of ERK and p38-MAPK, NF-kB activity). Moreover, other studies showed an antioxidant effect of bilberry leaves extract in serum of male Wistar rats under a double challenge composed of HFD and STZ. Likewise, in a recent study from our group, long-term supplementation of BB juice boosted serum antioxidant activity in a healthy Wistar rats. Nevertheless, there has been some controversy on BB-derived polyphenols benefits on total antioxidant activity in both in vivo and human studies. Distinct BB doses, food matrices and/or alternative methodologies to assess total antioxidant status may foster such ambiguity.
In humans, previous studies indicated that BB can have immunomodulatory effects and reduce oxidative stress in adults with metabolic syndrome. However, the results so far are relatively inconclusive, possibly due to considerable inter-individual variation. In a 8-week RCT, Basu et al. found that circulating levels of oxidized LDL, MDA and hydroxynonenal (HNE) decreased in participants with metabolic syndrome that consumed 50 g of freeze-dried BB beverage (approximately 350 g fresh BB) daily, despite no significant changes in serum glucose levels, lipid profiles or inflammation biomarkers. In another RCT designed to evaluate the effects of 6-weeks intake of 45 g of freeze-dried BB per day in subjects with metabolic syndrome, a significant decrease in superoxide and total ROS contents were found in whole blood and monocytes, together with decreased circulatory inflammatory markers and reduced gene expression of TNFα, TLR4 and IL-6 in monocytes. Similarly, other researchers have found that BB reduced inflammatory markers as well as enhanced antioxidant activity in individuals with metabolic syndrome or diabetes.
## Hepatoprotective effects of bb
Apart from high antioxidant and anti-inflammatory properties, BB may exert antidiabetic effects by additional mechanisms. It has been unveiled that BB phytochemicals can provide beneficial effects in T2DM by interfering with energy homeostasis and nutrient metabolism in different tissues, including in the liver.
Several phytochemicals abundant in BB, such as anthocyanins, have been shown to display lipid-lowering actions through hepatic metabolism pathways, including the regulation of key hepatic enzymes involved in lipogenesis and fatty acid catabolism, as well as in carbohydrate metabolism. A previous in vitro study suggested that concentrate anthocyanins and other polyphenols from BB may improve glucose metabolism by repressing glucose production in H4IIE rat hepatocytes. Additionally, two BB methanolic extracts showed hypoglycemic effects in human non-tumor hepatic LO2 cells, an effect correlated with the increase in GLUT-2 and PPAR-γ expression as well as with the inhibition of relevant inflammatory pathways.
The hepatoprotective effects of the BB phytochemicals have been highlighted in several animal models of chemical-induced liver injurybut also in obesity and diabetes-associated chronic diseases, including in NAFLD, as summarized in. A recent study showed that nonacylated anthocyanins from bilberry reduced plasma glucose, lipids and BCAA level in obese diabetic ZDF rats, suggesting an improvement of insulin sensitivity and reduction in lipogenesis. In another study in diabetic mice, antidiabetic effects of dietary anthocyanin-rich bilberry extract (BBE) were associated with targeting of hepatic AMPK, GLUT-4 and metabolic enzymes; furthermore, BBE downregulated the mRNA expression of two gluconeogenic enzymes, phoshoenolpyruvate carboxykinase (PEPCK) and glucose-6-phosphatase (G6Pase), and suppressed glucose flux into the blood. In addition, an enhanced phosphorylation of acetyl-CoA carboxylase (ACC), the key enzyme for FA synthesis that is downstream of AMPK, and the upregulation of PPAR-α, acyl-CoA oxidase (ACO) and CPT1, were found in the liver of diabetic mice supplemented with BBE. Furthermore, supplementation of BB polyphenols extracts also decreased the hepatic mRNA expression of SREBP-1 and fatty acid synthase (FAS), which could contribute to improved hepatic lipid metabolism. Other recent findings suggest that phenolic BB extract improved hepatic lipid metabolism via pathways involving the bile acids receptors farnesoid X receptor (FXR) and Takeda G protein-coupled receptor 5 (TGR5).
Moreover, BB phytochemicals elicit mitochondrial-targeted protective properties by scavenging ROS and acting as uncouplers of the OXPHOS, preserving mitochondrial function and protecting cells from apoptosis. Likewise, a protective effect of BB anthocyanin-rich extract was found in mice hepatic mitochondria against acrylamide-induced mitochondrial oxidative stress by inhibition of ROS formation. At the cellular level, C3G (one of the main components of BAE) reduced ROS formation, caspase-3 and -9 inactivation, and downregulated the pro-apoptotic Bax protein induced by hyperglycemia, preventing mitochondrial dysfunction through modulation of PI3K/AKT and JNK signaling pathways. The combination of BB juice and probiotics (BJP) reduced NAFLD-induced mitochondrial ultrastructure damage and swelling and hepatic necrosis in parallel with improvements on respiratory function. Moreover, BJP attenuated mitochondrial oxidative stress through elevation of GSH and SOD levels while reducing ROS production in an animal model of NAFLD. The authors demonstrated that the modulation of SIRT1/PGC-1α pathway is a potential target of BJP against hepatic damage induced by NAFLD. More recently, the activation of the AMPK/PGC-1α/SIRT3 signaling pathway was associated with beneficial effects of BB leaf polyphenols on hepatic mitochondrial function and oxidative defense. Additionally, improvements on mitochondrial respiratory parameters were found in isolated hepatic mitochondria from non-obese diabetic Goto-Kakizaki (GK) rats after drinking bilberry leaves decoction for 4 weeks. Even though a mechanistic explanation was not scrutinized in this study, quercetin present in bilberry leaves may partially explain the increased mitochondrial oxidative and phosphorylative activities in GK treated rats. In fact, this polyphenol has a positive impact on mitochondrial biogenesis related with an increased mRNA expression of PGC-1α and SIRT1 along with higher contents of mtDNA and cytochrome c. Remarkably, recent data from our group revealed that long-term BB fruits intake in a juice form elicited an accentuated bioenergetic remodeling in isolated hepatic mitochondria along with a metabolic transcriptional adaptive response in healthy animals. Further animal research also showed that the attenuation of hepatic steatosis and enhancement of lipolysis following exposure of hepatic cells to BB polyphenols can be promoted by modulation of autophagy.
In vivo animal studies have showed that BB reduced the aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels in both mouse models of liver injury but also in obese and insulin resistant animal models. Conversely, despite changes in lipid peroxidation in the liver, the hepatic TGs accumulation and serum AST and ALT levels were not altered following the administration of whole blueberry powder or polyphenolicrich extracts in diet-induced obesity and insulin resistant mice. Interestingly, authors found that the improved metabolic outcomes were only exhibited in obese insulin resistant animals supplemented with BB polyphenol-rich fractions and no improvements were found in mice fed high-fat high-sucrose diet plus whole BB powder, suggesting that the metabolic benefits are a result of specific classes of polyphenols isolated from the whole fruit.
In humans, some interventional studies have demonstrated the ability of BB to lower hepatic transaminases (such as ALT and AST), which might suggest hepatoprotection. Moreover, a double-blind RCT conducted in patients with NAFLD for 12 weeks demonstrated beneficial effects of purified anthocyanins derived from bilberry and black currant by improving insulin resistance. ↓ BW, liver and adipose tissue weight ↓ Insulin levels ↓ HOMA-IR ↓ TGs and T-Chol ↓ Hepatic steatosis ↓ Serum ceramide and DAG levels ↓ Serum PP2A and Prkcz mRNA expression ↓ ALT and AST activitiesObese Zucker rats Freeze-dried wild BB powder (8% of diet) 8 weeks = BW ↓ Fasting plasma TAG and T-Chol ↓ Hepatic SREBP-1 and FAS mRNA expression = Hepatic PPAR-α and PPAR-γ mRNA expression↓ Fasting plasma TNF-α, IL-6, and CRP ↑ Plasma adiponectin ↓ Hepatic TNF-α, IL-6, CRP and NF-kB mRNA expressionHFD-fed Sprague Dawley rat BB juice (15 g/kg) combined with probiotics 8 weeks ↓ Serum TGs, T-Chol, LDL-c and MDA ↓ Hepatic lipid accumulation ↑ Hepatic SIRT1and PPAR-α mRNA and protein expression ↓ Hepatic mRNA levels of SREBP-1c ↑ Serum SOD and GSH activities ↓ Serum ALT and AST activities↑ Hepatic mitochondrial swelling ↓ State 3 and 4 respiration rates ↑ RCR and ADP/O ratio ↑ Hepatic mitochondrial GSH and SOD mRNA and protein expression levels ↑ Hepatic protein and mRNA expression of SIRT1 ↑ Hepatic mRNA expression of PGC-1α ↓ Hepatic mitochondrial MDA levels and ROS activity↓ BW ↓ Serum and hepatic lipid and MDA levels ↑ Hepatic SOD and GPx activities ↓ TNF-α, IL-6 and NF-kB mRNA expression ↑ Fecal SCFAS Hepatic glycerophospholipids and glutathione metabolism and insulin-signaling pathways modulationDiabetic male KK-Ay mice BB anthocyanin extract (10 g/Kg diet) 4 weeks = BW and ↓ liver weight ↓ Serum glucose levels ↑ Insulin sensitivity ↓ Serum TGs and TC-Chol ↓ HGP ↑ Hepatic AMPK phosphorylation ↓ Hepatic lipid content ↓ Hepatic gluconeogenesis (↓ PEPCK and G6Pase mRNA expression) ↑ FAO (↑ PPAR-α, ACO and CPT1 mRNA expression)HFD-fed C57BL/6J mice Fermented BB juice 17 weeks ↑ BW gain ↑ Insulin sensitivity and glucose homeostasis ↓ Fasting serum glucose ↓ Serum LDL-c ↓ Serum TNF-α and ↑ IL-10 ↓ Hepatic fat accumulation ↑ Hepatic mRNA expression of IR and IRS ↓ Hepatic SCD1, SREBP1c and FAS expression ↑ mRNA and protein expression of GLUT-1; GCK, LDL-receptor and PPAR-α ↓ mRNA and protein expression of PPAR-γ ↑ Akt phosphorylationGK rats
Decoctions of Vaccinium myrtillus L. (bilberry) leaves 4 weeks = BW ↑ Glucose tolerance ↓ Occasional glycaemia ↑ Mitochondrial oxidative and phosphorylative activities ↑ State 3 respiration, FCCP-uncoupled respiratory and RCR activitiesACO, acyl-CoA oxidase; Akt, protein kinase B; ALT, alanine aminotransferase; AMPK, 5 adenosine monophosphate-activated protein kinase; AST, aspartate aminotransferase; BW, body weight; CPT1, carnitinepalmitoyltransferase-1A; DAG, diacylglycerol; FAO, fatty acid oxidation; FAS, fatty acid synthase; FCCP, carbonyl cyanide p-(tri-fluromethoxy)phenyl-hydrazone; G6Pase, glucose 6-phosphatase; GCK, Glucokinase; GLUT-1, glucose transporter 1; GPx, glutathione peroxidase; GK, Goto-Kakizaki; HGP, hepatic glucose production; HFD, high-fat diet; HOMA-IR, homeostatic model assessment for insulin resistance; IL, interleukin; LCHAD, long-chain hydroxyacyl-CoA dehydrogenase; LDL-c, low-density lipoprotein-cholesterol; MDA, malondialdehyde; PEPCK, phosphoenolpyruvate carboxykinase; RCR, respiratory control ratio; SCD1, stearoyl-CoA desaturase-1; SCFAS, short-chain fatty acids; SOD, superoxide dismutase; SREBP-1, sterol regulatory element-binding protein-1; TNF-α, tumor necrosis factor-α; T-Chol, total-cholesterol; ↑, increased; ↓, decreased; =, unchanged.
## Prebiotic-like effects of bb
Given that BB phytochemical composition comprises several bioactive compounds including high molecular weight PP and easily fermentable dietary fibers that are not well absorbed in the small intestine and reach the large intestine, such bioactive components are available to interact with GM. Indeed, prebiotic-like properties and/or antimicrobial effects against pathogenic intestinal bacteria of BB and their bioactive compounds have been reported in in vitro assays, in animal modelsand in human intervention studies, as outlined in. An in vitro study reported distinct positive correlations between different PP contents and bacterial species; in fact, while PP from BB bagasse powder was correlated with Actinobacteria (Bifidobacterium and Colllisella) and Akkermansia, the fiber content was associated with Faecalibacterium and Bifidobacterium. Moreover, the presence of anthocyanins and fibers in BB pomace promoted the growth of Lactobacillus and Ruminococcaceae genus and was negatively associated with Streptococcus. The addition of a BB extract to mixed human fecal bacterial populations enhanced the population size of Lactobacilli and Bifidobacteria.
Growing evidence hint that the positive effects of BB in improving glucose metabolism and energy balance are closely linked with their ability to modulate the composition and/or function of GM.summarizes the most relevant studies performed in animal models. A recent study reported that upon intestinal absorption BB polyphenols improved metabolic health in diet-induced obese mice via modulation of GM rather than through systemic effects. It was also observed that germ-free mice receiving fecal microbiota transplantation (FMT) from mice fed BB polyphenolic fractions (specifically enriched with proanthocyanins or anthocyanidins) were protected from diet-induced obesity and presented an improved insulin sensitivity and glucose homeostasis when compared to mice receiving GM from HFHS fed mice. Furthermore, authors postulated that the beneficial effects of FMT could be the result of a decreased proportion of BCAA in the feces of mice fed whole BB or just the polyphenols. Moreover, positive effects on insulin sensitivity and glycemia were found in the mouse model of HFD-induced obesity upon BB juice supplementation for 17 weeks, with or without fermentation; the authors hypothesized that this protection was closely related with BB ability to stimulate SCFAs production and maintain GM homeostasis. Among health commensal bacteria, Akkermansia muciniphila (AM) has been positively correlated with enhanced mucosal barrier function and improved metabolic phenotype in obesity ad T2DM. Notably, BB interventions enhanced AM levels in animal models of metabolic syndrome. Indeed, whilst favouring the growth of healthier bacteria, BB constituents seem to display the ability to restore gastrointestinal integrity by: (i) enhancing colonic mucus thickness; (ii) increasing the expression of tight junctions proteins; (iii) overexpressing MUC-2 mRNA, the primary glycoprotein of the gastrointestinal mucus layer; (iv) reducing endotoxemia and intestinal inflammationand (v) fortifying the mucosal immunity.
Interestingly, BB polymeric proanthocyanidins have been recently shown capacity to restore the colonic mucus layer, modulate GM and attenuate glucose tolerance in obese mice, in contrast to the anthocyanin fraction from BB. However, other subclasses of polyphenols, secondary metabolites and fibers also present in BB may potentially account for these prebiotic effects counteracting the metabolic disorders. Notwithstanding, ambiguous studies and conflicting results still prevail, which could possibly be explained by several factors such as distinct duration and mode of administration of BB interventions, differences in the BB phytochemical composition (i.e., the type and structure of polyphenolic constituents), as well as differences in the dietary regimens. Noteworthy, the inclusion of bilberries in the high-fat setting prompted an increase in SCFA contents accompanied by the normalized α-diversity of GM, enhanced the abundance of Bacteroidetes, Bifidobacterium and butyrate-producing bacteria, which collectively contributed to thwart pre-obesity events like hepatic lipid accumulation.
The impact of BB on GM has been also evaluated in human interventional studies. Vendrame et al.showed that 6 weeks consumption of a BB powder drink rich in anthocyanins significantly enhanced the abundance of Bifidobacteria in the cecum from healthy volunteers, whereas no differences were observed for Bacteroides spp., Prevotella spp., Enterococcus spp., and Clostridium coccoides. A pilot RCT was carried out in overweight and obese subjects to evaluate the effects of a gastrointestinal microbiome modulator (GIMM), composed of inulin, β-glucan and anthocyanins and polyphenols from BB, on metabolic parameters, fecal markers of GM, and satiety. Thirty overweight or obese participants consumed the GIMM or placebo, daily, for 4 weeks. Improved glucose tolerance restored satiety hormones (increased PYY and decreased ghrelin) and fecal SCFAs contents were observed in the GIMM-treated group when compared to the placebo; however, no significant changes were found between groups regarding insulin sensitivity, fecal markers of GM, plasma satiety hormones or serum lipid concentrations. The prebiotic effects of BB reported in the preclinical ground need further validation in the clinical setting, namely in the (pre)diabetic state, as the literature is particularly scarce in human interventions in this condition. There is one ongoing clinical trial (NCT03266055) exploring the prebiotic effects of BB in subjects with metabolic syndrome. The results from this and other trials further conducted will be decisive to achieve more robust conclusions regarding the use of BB to prevent dysbiosis in prediabetes.
## Concludings
Herein, we provide a review of the current knowledge in this field by revisiting preclinical data obtained from in vitro assays and animal models together with studies from human subjects, to disclose the mechanisms by which BB may be a valuable approach in prediabetes. Through improving glycemic control, reducing oxidative stress and inflammation, regulating TG's accumulation and trafficking, along with ameliorating GM dysbiosis, this multi-modal functional fruit positively impacts a diversity of pathophysiological mechanisms that occur as early as prediabetes and for which pharmacological interventions (either oral antidiabetics or weight-loss drugs) are not approved in practical guidelinesor are only considered in cases at high risk. In this sense, BB have been shown to modulate hepatic and intestinal molecular targets of currently approved antidiabetics [α-glucosidase inhibitors (e.g., acarbose), DPP-4 inhibitors (e.g., sitagliptin), PT1B inhibitors (rosiglitazone)], whose benefits often surpass the glucocentric view of diabetes management. As a matter of fact, BB benefits also goes beyond the improvement of glycemic control as they positively modulate a panoply of metabolic functions that ensues not only in the pancreas but also in the gut and the liver, two pathophysiologic paths that critically determine (pre)diabetes progression. Collectively, it is outlined how blueberries may provide a fruitful source of phytochemicals able to prevent (pre)diabetes progression.
As final critical analysis, two serious limitations narrow our current understanding of BB health benefits on (pre)diabetes. Firstly, many in vivo and in vitro studies were con-ducted with BB-derived PP in supraphysiological doses, unlikely to be achieved upon human daily consumption. Secondly, most in vitro studies use undigested dietary BB-derived polyphenols and miss the corresponding in vivo metabolites/conjugates. Collectively, these data have a low predictive value. Finally, RCTs greatly diverge on BB supplementation regimens. Future research should focus on the qualitative and quantitative evaluation of BB polyphenols in both systemic circulation and target tissues. Moreover, longer-term RCTs with standardized interventions (e.g., doses employed, duration) and readouts will hopefully bring stronger evidence to guide a more effective use of this functional fruit with remarkable antioxidant activity in the early stage of diabetes.
## Conflicts of interest:
The authors declare no conflict of interest. |
An enhanced variant effect predictor based on a deep generative model and the Born-Again Networks
The development of an accurate and reliable variant effect prediction tool is important for research in human genetic diseases. A large number of predictors have been developed towards thisgoal, yet many of these predictors suffer from the problem of data circularity. Here we present MTBAN (Mutation effect predictor using the Temporal convolutional network and the Born-Again Networks), a method for predicting the deleteriousness of variants. We apply a form of knowledge distillation technique known as the Born-Again Networks (BAN) to a previously developed deep autoregressive generative model, mutationTCN, to achieve an improved performance in variant effect prediction. As the model is fully unsupervised and trained only on the evolutionarily related sequences of a protein, it does not suffer from the problem of data circularity which is common across supervised predictors. When evaluated on a test dataset consisting of deleterious and benign human protein variants, MTBAN shows an outstanding predictive ability compared to other well-known variant effect predictors. We also offer a user-friendly web server to predict variant effects using MTBAN, freely accessible at http:// mtban. kaist. ac. kr. To our knowledge, MTBAN is the first variant effect prediction tool based on a deep generative model that provides a user-friendly web server for the prediction of deleteriousness of variants.OPENshows superior predictive performances compared to other variant effect predictors. Our model is fully unsupervised and is not dependent on labelled data for training. This gives the model advantage over supervised predictors, for which data circularity is an inherent problem. We also offer a freely accessible web server for using MTBAN for variant effect prediction.MethodsMTBAN model. We previously developed a deep autoregressive generative model for variant effect prediction, called mutationTCN 4 . For each protein variant, the model is trained on the multiple sequence alignment of the corresponding protein. As it is a generative model, it is trained by maximizing the likelihood of the training data, which is equivalent to minimizing the negative log likelihood between the input sequence and the predicted output. After training, the model can predict the probability of observing a given protein sequence under the parameters of the trained model. The deep autoregressive generative model is implemented using the temporal convolutional network architecture 6 . Each sequence from the input multiple sequence alignment is encoded by representing each amino acid in the sequence as a distinct integer. The input is passed through an embedding layer, followed by a series of dilated causal convolution layers, an attention layer, and a fully connected layer(Fig. 1). We showed that this model can effectively capture information from evolutionarily related sequences and use this information to predict the functional effects of variations in a sequence 4 . MTBAN combines this model with a knowledge distillation technique in machine learning, known as the Born-Again Networks (BAN) 5 . Knowledge distillation is a process of transferring the knowledge from one machine learning model to another 7 . In this scheme, the former is referred to as the "teacher" model and the latter is referred to as the "student" model. Typically, knowledge is transferred from a larger model to a smaller model, which allows for the reduction of model size while maintaining similar predictive power as the original model. In the setting of BAN, the student network is of the same capacity as the teacher network, which enables the student network to outperform the teacher network 5 . We found that the BAN framework in which both the teacher and the student network is implemented with mutationTCN outperforms the original mutationTCN model.The model structure of MTBAN is shown inFig. 1. In the first step, only the teacher network is trained, with the loss function being the label loss, which refers to the cross entropy loss between the input sequence and the softmax output distribution of the teacher network. In the next step, only the student network is trained, with the loss being the sum of the label loss and the teacher loss. Here, the label loss refers to the cross entropy loss between the input sequence and the softmax output of the student network. The teacher loss refers to the cross entropy loss between the softmax output of the student network and the softmax output of the teacher network.
While recent sequencing technologies have resulted in a tremendous amount of sequence variant data, the identification of deleterious variants is still a difficult problem. Development of a reliable computational tool to predict the effects of sequence variants would aid in the treatment of many human genetic diseases. To achieve this goal, many predictors have been developed based on different approaches. Among these methods, supervised methods learn from labelled variant data consisting of known deleterious and benign variants, and many of them show good predictive ability. However, many supervised methods face the problem of data circularity, which can be divided into two types according to Grimm et al. [bib_ref] The evaluation of tools used to predict the impact of missense variants..., Grimm [/bib_ref] The type I circularity arises due to the overlap between training data and test data. The type II circularity occurs when all variants in a given gene are labelled either all deleterious or all benign, which results in the model predicting the same label for all variants in that gene. Previous studies [bib_ref] The evaluation of tools used to predict the impact of missense variants..., Grimm [/bib_ref] [bib_ref] Variant effect prediction tools assessed using independent, functional assay-based datasets: Implications for..., Mahmood [/bib_ref] [bib_ref] Using deep mutational scanning to benchmark variant effect predictors and identify disease..., Livesey [/bib_ref] have suggested that this problem of data circularity can result in an inflation of the reported performances of many supervised predictors. On the other hand, unsupervised methods do not learn from labelled variant data and learn solely from the evolutionary information contained in multiple sequence alignments. A recent study which carried out an extensive comparison of variant effect predictors claimed that a class of unsupervised models, namely the deep generative model, is a promising area of research for variant effect prediction [bib_ref] Using deep mutational scanning to benchmark variant effect predictors and identify disease..., Livesey [/bib_ref].
Here, we introduce MTBAN (Mutation effect predictor using the Temporal convolutional network and the Born-Again Networks), an enhanced method to predict the deleteriousness of single amino acid variants. We previously developed a method called mutationTCN 4 based on a deep autoregressive generative model, and showed that it demonstrates state-of-the-art performances on the prediction of functional effects of variants. In this work, we apply a knowledge distillation technique called the Born-Again Networks (BAN)to the muta-tionTCN model and develop an improved model called MTBAN. In machine learning, knowledge distillation is a process involving the transfer of knowledge learned from one machine learning model to another. Using the Born-Again Networks allows the student network to achieve an improved predictive power compared to the teacher network. When evaluated on human variant datasets with deleterious and benign variants, MTBAN
[formula] p i = exp( z i T ) j exp( z j T ) Figure 1. MTBAN model structure. [/formula]
We implemented BAN with mutationTCN as both the teacher and the student network. In the first step, only the teacher network is trained, with the loss function being the label loss (red arrow), which refers to the cross entropy loss between the input sequence and the softmax output distribution of the teacher network. In the second step, only the student network is trained, with the loss being the sum of the label loss (red arrow) and the teacher loss (blue arrow). Here, the label loss refers to the cross entropy loss between the input sequence and the softmax output of the student network. The teacher loss refers to the cross entropy loss between the softmax output of the student network and the "softened" output distribution of the teacher network. www.nature.com/scientificreports/ where z i is the logit computed for each class and T is the temperature parameter, which is typically set to 1 7 . Using higher temperatures leads to more "softened" output distributions. According to Hinton et al., these softened output distributions contain the "dark knowledge," which is the hidden knowledge learned by the teacher network. In BAN, the transfer of this "dark knowledge" from the teacher to the student contributes to the improved performance of the student network. In our implementation, we used a temperature of 4. We trained both teacher and student networks for 500,000 iterations using the mini-batches with the size of 128. For both teacher and student networks, the learning rate is set to 0.001 when the number of training iterations is smaller than 3000, and 0.0001 when it is greater than 3000.
We computed the predictions of MTBAN for a total of 1605 human protein alignments provided by Hopf et al. [bib_ref] Mutation effects predicted from sequence co-variation, Hopf [/bib_ref] These pre-computed predictions on the Hopf dataset were used for evaluating the model on the test set. According to Hopf et al., their alignment generation protocol involves multiple iterations of profile HMM homology search in an attempt to ensure that there are enough sequences in the alignment and that the alignment coverage of the target protein sequence domain is sufficient [bib_ref] Mutation effects predicted from sequence co-variation, Hopf [/bib_ref]. This allows us to obtain an alignment that contains as much evolutionary information as possible.
Model outputs. For a given variant, the model outputs the log probability score, the z-score, the probability of deleteriousness, and the predicted label. First, the log probability score is given by the following:
where p(x mutant |θ) and p(x wild-type |θ) are the probability assigned to the mutant sequence and the wild-type sequence, respectively, by the generative model with parameters θ . The log probability score is easily computed from the loss function, as the model loss function is the negative log likelihood 4 . The smaller the score, the more likely the variant has a deleterious effect. Second, the z-score is computed by normalizing the distribution of log probability scores for all possible missense variants against the target protein sequence. This normalization process is done due to the variations in the score distributions across different proteins. Third, the probability of deleteriousness for each variant, ranging from 0 to 1, is computed. This is determined from the set of variants in the Humsavar database (release 03/2020) 9 which overlap with our pre-computed model predictions for the Hopf dataset, which are 1221 deleterious and 1221 benign variants. We obtained the z-score distribution for this set of variants, divided the distribution into equal-length z-score intervals, and calculated the proportion of deleterious variants in each z-score interval. Finally, using the same z-score intervals, we determined a z-score cutoff which maximizes the classification accuracy . This cutoff is used to assign a predicted label, either deleterious or benign, to a given variant.
## Evaluation datasets.
To evaluate the ability of the model to classify human protein variants as deleterious or benign, we created a test dataset by combining the variants from the datasets used by Grimm et al. [bib_ref] The evaluation of tools used to predict the impact of missense variants..., Grimm [/bib_ref] and Mahmood et al. [bib_ref] Variant effect prediction tools assessed using independent, functional assay-based datasets: Implications for..., Mahmood [/bib_ref] Details regarding the datasets can be found in . We used the HumVar dataset from Grimm et al., which contains human protein variants that are known to be disease-causing or neutral 1 . Also, we used the UniFun, BRCA1-DMS, and TP53-TA datasets from Mahmood et al., which contain deleterious and benign protein variants determined from direct in vitro functional assays, such as the deep mutational scanning experiment 2 . Mahmood et al. pointed out that commonly used disease-related variant datasets often overlap with the training data used by supervised predictors 2 . Because of this reason, they created the functionally determined variant datasets in order to avoid the problem of data circularity and establish an independent test set for benchmarking 2 . Another study 3 also supports this claim and uses the data from deep mutational scanning experiments to benchmark a large number of variant effect predictors. Also, it is reported that the Critical Assessment of Genome Interpretation (CAGI), which aims to perform an unbiased assessment of variant effect predictors, uses data from deep mutational scanning experiments as part of their benchmark dataset [bib_ref] Reports from CAGI: The critical assessment of genome interpretation, Hoskins [/bib_ref] [bib_ref] Identifying a high fraction of the human genome to be under selective..., Davydov [/bib_ref] , and phyloP [bib_ref] Detection of nonneutral substitution rates on mammalian phylogenies, Pollard [/bib_ref]. The predictions of the commonly used predictors on the test dataset were obtained from dbNSFP 21 via the Ensembl variant effect predictor [bib_ref] The ensembl variant effect predictor, Mclaren [/bib_ref]. Since the score cutoffs for phyloP, DANN, phastCons, GERP++, MPC, and GenoCanyon were not provided by dbNSFP, we computed the cutoffs for each predictor using the Humsavar database (release 03/2021) as described in "Methods" section.
We found variants among the datasets from Grimm et al. and Mahmood et al. for which MTBAN predictions exist in the pre-computed Hopf dataset, and used those variants for comparison with other methods. Since the number of deleterious variants was significantly larger than that of benign variants, we randomly selected variants from the deleterious variant data to match the data size of the deleterious variants and the benign variants. This resulted in a balanced test set consisting of 1709 deleterious and 1709 benign variants in total.
Evaluation criteria. The following metrics were used for evaluating the classification ability of the variant effect predictors: ROC-AUC (Receiver Operating Characteristic Area Under Curve), PR-AUC (Precision-Recall Area Under Curve), accuracy, Matthews Correlation Coefficient (MCC), precision, specificity, sensitivity, F-score, and Negative Predictive Value (NPV). For MTBAN, ROC-AUC and PR-AUC were calculated using z-scores, and other evaluation metrics were calculated using the predicted label. The following equations were used for computing the evaluation metrics:
where TP, TN, FP, and FN are the number of true positives, true negatives, false positives, and false negatives, respectively.
# Results
## Evaluation on human protein variant datasets.
We assessed MTBAN and other variant effect predictors on the task of classifying human protein variants as deleterious or benign. As described in "Methods" section, our test dataset combines the disease-associated variants from Grimm et al. [bib_ref] The evaluation of tools used to predict the impact of missense variants..., Grimm [/bib_ref] and functionally determined variants from Mahmood et al. [bib_ref] Variant effect prediction tools assessed using independent, functional assay-based datasets: Implications for..., Mahmood [/bib_ref] , resulting in a total of 1709 deleterious and 1709 benign variants. When compared with 11 other variant effect predictors in terms of ROC-AUC and PR-AUC, our model outperformed all other predictors, achieving a ROC-AUC of 0.883 and a PR-AUC of 0.878 [fig_ref] Figure 2: ROC Curves and Precision-Recall Curves for MTBAN and other predictors on the... [/fig_ref]. Even though our model is fully unsupervised, its predictive ability outperforms the supervised predictors including PolyPhen-2, whose training dataset has overlapping variants with the dataset from Grimm et al. [bib_ref] The evaluation of tools used to predict the impact of missense variants..., Grimm [/bib_ref] Also, MTBAN achieved the highest accuracy, MCC, and F-score among all compared variant effect predictors. In addition, our model demonstrates a good balance between specificity and sensitivity, unlike fathmm-MKL or phyloP which demonstrate good performance in only one of the two measures.
In addition, we conducted further assessment using only the disease-associated variant data from Grimm et al. [bib_ref] The evaluation of tools used to predict the impact of missense variants..., Grimm [/bib_ref] , and using only the functionally determined variant data from Mahmood et al. [bib_ref] Variant effect prediction tools assessed using independent, functional assay-based datasets: Implications for..., Mahmood [/bib_ref] When tested on the data from Grimm et al. consisting of 1230 deleterious and 1230 benign variants, our model achieved the highest ROC-AUC, PR-AUC, accuracy, MCC, and F-score . Also, when tested on the data from Mahmood et al. consisting of 479 deleterious and 479 benign variants, our model achieved the highest ROC-AUC, accuracy, MCC, and F-score [fig_ref] Table 2: Performances [/fig_ref]. Overall, MTBAN shows an outstanding classification ability in both disease-associated variant data and functional assay-derived variant data.
Web server. We offer a user-friendly web server which predicts variant effects using MTBAN (Supplementary [fig_ref] Figure 2: ROC Curves and Precision-Recall Curves for MTBAN and other predictors on the... [/fig_ref]. The server takes in as input a protein UniProt accession and a list of amino acid variants. Upon receiving input, it determines the target protein sequence region, and checks if pre-computed predictions exist www.nature.com/scientificreports/ for the given variants. If they exist, the server immediately returns predictions to the user. Otherwise, it checks if a multiple sequence alignment of the target protein sequence region is present in the database. If an alignment is present, it uses that alignment for subsequent computations. If an alignment is not present, it generates one using a profile HMM homology search tool [bib_ref] Accelerated profile HMM searches, Eddy [/bib_ref] and saves it in the database. During the computation, alignment columns that have more than 30% gaps are dropped. If some of the input variants belong to these un-aligned columns in the alignment, those variants are excluded from prediction and are indicated in the results. The next step is the computation of sequence weights, based on the similarity of sequences in the alignment. This step is included to reduce any sequence bias present in the multiple sequence alignment 4 . Afterwards, the prediction model is trained, and the server returns predictions to the user. After job processing, the predictions are saved so that the server can immediately return the results when the same set of mutations are later submitted as input. In the web server implementation, due to time constraints, the MTBAN teacher network and student network are both trained for 200,000 iterations, with learning rate 0.001.
# Discussion
Here, we have introduced MTBAN, an improved method for predicting the deleteriousness of single amino acid variants. As demonstrated in our previous work 4 , the deep autoregressive generative model is a powerful tool for learning the distribution underlying the evolutionarily related sequences of a protein and predicting the effects of variations in a sequence. Combining the deep autoregressive generative model with a knowledge distillation method known as the Born-Again Networks (BAN) further improves the predictive power of the model, by transferring the knowledge learned by the model to the second model of the same capacity. We conducted an assessment using the test set combining the disease-related variants from Grimm et al. [bib_ref] The evaluation of tools used to predict the impact of missense variants..., Grimm [/bib_ref] and the functionally determined variants from Mahmood et al. [bib_ref] Variant effect prediction tools assessed using independent, functional assay-based datasets: Implications for..., Mahmood [/bib_ref] , and further assessment using each of the two variant sets. In all cases, MTBAN consistently shows outstanding predictive ability compared to other prediction tools. The results indicate that MTBAN is a reliable method for predicting the deleteriousness of human protein variants. Previous works 1-3 have pointed out concerns regarding the problem of data circularity in many supervised predictors, which can lead to an inflation of the reported performances of these tools. Due to the fully unsupervised nature of MTBAN, it is not hindered by the problem of data circularity and can be considered to have higher generality compared to supervised models. Moreover, while we only considered human protein variants in this work, it is possible to predict the effects of protein variants in any other species if a multiple sequence alignment is available.
As previously mentioned, the BAN involves the transfer of the "dark knowledge" hidden in the softened output distribution of the teacher network to the student network. We speculate that due to the large size and the high complexity of the training set used in this study, the student equipped with the teacher's knowledge can better model the distribution of the training data, compared to the teacher alone. In other scenarios where the model is of high capacity and the training data is limited in size, the student network may possibly perform worse due to overfitting.
One potential limitation of MTBAN and mutationTCN is that they can only make predictions for variants which correspond to the conserved positions in the multiple sequence alignment of a protein. However, when we analyzed all of the 9935 human protein multiple sequence alignments in the Hopf dataset, approximately 88% of the target sequences were conserved, which is a considerably large proportion. Another potential limitation of MTBAN is that the training time is longer compared to mutationTCN alone for prediction. Although MTBAN takes a longer time to train, it shows a higher predictive performance compared to the previous model.
The results of our work show that the deep generative model is a powerful tool for predicting the effects of sequence variations. We expect that deep generative models will continue to play an important role in discovering the effects of genetic variants. In addition, to our knowledge, MTBAN is the first variant effect prediction tool based on a deep generative model that provides a user-friendly web server for the prediction of deleteriousness of variants. This method is expected to be a useful tool for the prioritization and identification of variants involved in human genetic diseases.
## Data availability
The datasets generated during and/or analyzed during the current study are available at https:// github. com/ ha019 94/ MTBAN.
[fig] Figure 2: ROC Curves and Precision-Recall Curves for MTBAN and other predictors on the test dataset. (a) MTBAN achieved a ROC-AUC (Receiver Operating Characteristic Area Under Curve) of 0.883, which is the highest among 12 variant effect predictors. (b) MTBAN achieved a PR-AUC (Precision-Recall Area Under Curve) of 0.878, outperforming all other variant effect predictors. [/fig]
[table] Table 2: Performances [/table]
|
Health morbidity in Brazilian prisons: a time trends study from national databases
# Abstract
Objective The goal of this study was to explore the surveillance data about mandatory reporting diseases, included in the official information systems, and evaluate the historical trend analysis in prisoners in Design A time trends study was performed using secondary data from prisons' health units. From a total of 27 states, 12 of them presented a growth in disease notifications, 14 were stable and in only one state was there a decrease in notifications. There was an increase in notifications in the country as a whole. Tuberculosis (64.4%), dengue (9.1%), AIDS (9.0%) and viral hepatitis (5.9%) were among the most frequently reported diseases during the study period. Conclusion Despite showing stable tendencies, our results show high rates of diseases in Brazilian prisons. Prison health services should not be isolated but integrated into regional and national health and justice systems.
# Introduction
The United Nations Standard Minimum Rules for the Treatment of Prisoners, known as the Nelson Mandela Rules, is a resolution adopted by the General Assembly on 17 December 2015, which recommends regarding healthcare services that: (a) The provision of healthcare for prisoners is a State responsibility. Prisoners should enjoy the same standards of healthcare that are available in the community, and should have access to necessary healthcare services free of charge without discrimination based on their legal status; and (b) Healthcare services should be organised in close relationship with the general public health administration in a way that ensures continuity of treatment and care, including HIV, tuberculosis and other infectious diseases as well as drug dependence.In other words, the fact that individuals are in prison does not mean that they have reduced rights to adequate healthcare. Despite finding themselves deprived of their liberty, prisoners are still entitled to have their fundamental rights protected and the prison system must have the proper infrastructure to offer them rights such as health, education, work and other needs that in general should not be suppressed by the court's ruling. Those rights, if fulfilled, will contribute to the process of social rehabilitation. In Brazil, this paradox is aggravated by the adverse conditions found in prisons, such as the precariousness of physical spaces, scarcity of specialised human resources, lack of adequate healthcare and presence of structural violence practices. [bib_ref] The need to improve health care in prisons, Fernandes [/bib_ref] [bib_ref] Sanitation rights, public law litigation, and inequality: a case study from Brazil, De Barcellos [/bib_ref] In the specific case of health conditions, instead of being in an environment able to safeguard physical and mental health, distress is often faced with an unhygienic and unhealthy environment. It strengths and limitations of this study ► This study was nationwide representative and included data about mandatory reporting disease cases that were notified in prisons units in Brazil. ► These data presented an 8-year trend analysis of notified diseases in Brazilian prisons' units. ► The use of secondary data presented some limitations but it represented the best information where there was no other source of data. ► These data were carried out from all units identified as prison health facilities in Brazil.
## Open access
facilitates the spread of communicable diseases, injuries related to violence and mental disorders. Communicable diseases are frequently transmitted among prisoners, and the rates of HIV, viral hepatitis and tuberculosis are much higher and complicated to deal with when compared with the general population. [bib_ref] Imprisoned and imperiled: access to HIV and TB prevention and treatment, and..., Todrys [/bib_ref] [bib_ref] HIV and tuberculosis in prisons in sub-Saharan Africa, Telisinghe [/bib_ref] [bib_ref] Active and latent tuberculosis in Brazilian correctional facilities: a cross-sectional study, Carbone [/bib_ref] [bib_ref] A Cross-Sectional Survey of HIV Testing and Prevalence in Twelve Brazilian Correctional..., Sgarbi [/bib_ref] [bib_ref] Prevalence and Incidence of HCV Infection among Prisoners in Central Brazil, Puga [/bib_ref] [bib_ref] Harm Reduction, and Persons Deprived of Liberty: What Standards Does International Human..., Sander [/bib_ref] International and national studies also report a high prevalence of mental health problems, including substance abuse disorders, and a higher prevalence of non-communicable diseases among prisoners. [bib_ref] Substance use disorders in prisoners: an updated systematic review and meta-regression analysis..., Fazel [/bib_ref] [bib_ref] Disease prevalence and use of health care among a national sample of..., Rosen [/bib_ref] [bib_ref] Addiction Treatment Within U.S. Correctional Facilities: Bridging the Gap Between Current Practice..., Wakeman [/bib_ref] [bib_ref] Drug-related disorders and the criminal and clinical background of the prison population..., Mendes Dos [/bib_ref] Historically in Brazil, the concern regarding healthcare for prisoner populations has been low, fragmented and vertical; insofar as policies have been developed, they were limited to aspects that focused on isolated health problems and specific diseases. [bib_ref] The need to improve health care in prisons, Fernandes [/bib_ref] In 2014, the Brazilian Ministry of Health formulated the 'National Policy for Integral Attention to the Health of Incarcerated Persons in the Prison System' in order to organise health activities and services for prisons within the scope of the national primary care model.This initiative extended the effective coverage of the National Health System (In Portuguese Sistema Único de Saúde [SUS]) to this particularly vulnerable population, seeking to achieve the universality of the SUS while guaranteeing the constitutional right to health with equity and integrality for prisoners.
The imprisonment organisation in Brazil is a responsibility of each federal state, producing different profiles depending of the state administration. According to the most recent report of the National Prison Monitoring Database of the National Justice Council, [bib_ref] National Data Bank of Arrest Warrants -BNMP, Brazil [/bib_ref] the total population of inmates in Brazil is 630 092 inmates distributed in 1.423 state prisons units and four federal prisons units, 95% are men and 5% are women. Since the expansion of effective coverage of SUS for prisoners in Brazil, the epidemiological profile of this public health system, considered one of the largest in the world, has been alarming. Available data are scarce, fragmented and related to certain types of injuries, which makes difficult to understand the whole situation.
The only available information is about the mandatory reporting diseases that are in the official information systems from the Ministry of Health and the Ministry of Justice, but it is usually incomplete. The goal of this study was to explore the surveillance data about mandatory reporting diseases, included in the official information systems, and evaluate the historical trend analysis in prisoners from Brazil.
## Methods type of study
This study was a time trends evaluation, conducted using secondary data, from mandatory reporting disease cases from the incarcerated population in Brazil from 2007 to 2014. Diseases diagnosis and individual data were available at the National System of Disease Notification (in Portuguese SINAN), Mortality Information System (in Portuguese SIM) and Prison Registration Systems (in Portuguese INFOPEN and GEO prisons). The units identified as prison's health facilities in the National Health System sent the information to the official information systems.
## Patient and public involvement
Databases used in this study were obtained from the Ministry of Health, Ministry of Justice and the National Council of Justice. They will have access to the results of this study to be able to implement health access strategies for this population. Patients and public were not directly involved in the study.
## Variables
The following variables were studied: state of origin and year of notification (the majority of prisons facilities are administrated by states), gender, age, ethnicity, schooling and notified diseases following the International Classification of Diseases, version 10. In Brazil, the notified diseases are included in the list of mandatory reporting diseases. They are selected through criteria such as magnitude, dissemination potential, social transcendence, vulnerability, availability of control measures and international commitment with eradication programmes obliging and universalizing reporting. The reports are important for the rapid control of those events that require prompt intervention.After performing descriptive analysis, the following indicators were proposed:
# Data analysis
To gather epidemiological data on mandatory reporting diseases in the incarcerated population, analyses of SINAN notification data at the national level were performed. SINAN data were consolidated with SIM, INFOPEN and GEO prison data. Quantitative data were analysed using the Statistical Package for the Social Sciences (SPSS, V.20.0) and Stata statistical software, V.11.2. A descriptive analysis was performed including frequency distribution for qualitative variables. Figures were prepared from the historical series and linear regression models of the variables observed throughout the study years were estimated. In addition, we presented the regression model and the value of R 2 , which indicates the portion of the variability explained by the linear trend; p value of F-test was also estimated, as well as the classification of the trend as stable (no trend), increasing or decreasing. The level of significance was set at 5%.
## Open access ethical procedures
This study was carried out with authorisation from the Ministry of Health, which released the databases without disclosing individual identification data. The southeastern region is the geographic area with the largest number of prisoners in Brazil and that which reported the highest number of cases of disease notifications, a total of 13 149 (56.3%) cases; São Paulo state had the highest number of notifications. Subsequently, the south and northeast regions reported more cases, with 3231 (13.8%) and 3213 (13.8%) cases, respectively. In general, it was observed that the number of diseases notifications increased over the years, which may represent a greater access to health services. [fig_ref] Figure 2: Trends of compulsory notification diseases reported among prisons' population in Brazil, 2007-2014 [/fig_ref] shows the proportions of disease reports in Brazil per year, by the prison population per 1000 inhabitants. [fig_ref] Table 1: Linear regression model with the proportion of cases reported by geographical regions... [/fig_ref] presents the linear regression model with the proportion of cases reported by geographical regions and states of Brazil in the incarcerated population. The regression models were used to measure trends in the proportion of cases reported in each state and in the general total of Brazil between years 2007 and 2014. This table presents the model and the value of R 2 , which indicates the proportion of the variability explained by the linear trend. According to the results, it was observed that from the total of 27 states, 12 of them presented a growth in disease notifications, 10 were stable, only 1 presented a decrease in disease notifications and 4 states (one from the North and three from the Northeast Regions) did not notify any case of disease during the period. These four states present less disadvantaged economical conditions. For Brazil, in general, there was an increase in disease notifications.
Most mandatory reporting diseases are communicable diseases. Tuberculosis (70.1%), AIDS (7.5%), syphilis (5.6%), dengue (4.5%) and viral hepatitis (3.5%) are among the most frequently reported diseases in 2014 and Open access also during all the study period. It is also interesting to note the presence of antirabies vaccine (2.1%) among the frequent ones . [fig_ref] Table 3: Linear regression model of reported diseases proportion and its trends, 2007-2014 [/fig_ref] describes the results of the linear regression model to measure the trends in the proportion of cases reported between the years 2007 and 2014, according to the mandatory reporting disease. The cases notification increased in this period for syphilis and syphilis in pregnancy, male urethral discharge, Chagas disease, American cutaneous leishmaniasis and adverse effects after vaccination. Only the antirabies human vaccine demand and the exanthema lesions decreased during the study years.
# Discussion
It is the first study including nationwide representative health data of prisoners in Brazil. The main contribution of this study was to identify the frequency of mandatory reporting diseases in Brazilian prisons and its trends. The mandatory reporting of diseases increased in the majority of Brazilian states and main notified diseases were stable during the study period, independent of the geographical region. In general, these results can be explained by improvements in the surveillance system and because of the presence of healthcare teams inside prisons.
Tuberculosis, AIDS, dengue fever and viral hepatitis were the most frequent notified diseases in the period. These data are in agreement with other studies that show high rates of tuberculosis in Brazilian prisons. Regarding HIV, a study in the central-west region of São Paulo state found 1.6% of infection in 12 prisons. [bib_ref] A Cross-Sectional Survey of HIV Testing and Prevalence in Twelve Brazilian Correctional..., Sgarbi [/bib_ref] A literature review on HIV and tuberculosis in sub-Saharan African prisons identified data from only 24 of Open access the 49 countries in the region. In countries where data were available, they were frequently of low quality and hardly nationally representative. The prevalence of HIV infection ranged from 2.3% to 34.9%, and of tuberculosis from 0.4% to 16.3% in prisoners; they nearly always had a higher prevalence of both diseases than did the non-incarcerated population in the same country. 9 A wide-ranging literature review, including data published between 2005 and 2015, aimed to understand the global epidemiology of these infections in prisoners and found that 3.8% had HIV, 15.1% had hepatitis C virus (HCV), 4.8% had chronic hepatitis B virus (HBV) and 2.8% had active tuberculosis. The authors found higher rates in prison populations when compared with general population, mainly because of the criminalisation of drug use and the detention of people who use drugs. [bib_ref] Global burden of HIV, viral hepatitis, and tuberculosis in prisoners and detainees, Dolan [/bib_ref] Syphilis and male urethral discharge were among the infections that presented increased notification rates in the study period. It is important to highlight that the number of syphilis cases among adults in Brazil has been consistently increasing, mainly in the vulnerable population, since 2010 23 . The situation is not different among prisoners in the country, who present even a higher rate of the infection. [bib_ref] High Prevalence of Treponema pallidum Infection in Brazilian Prisoners, Correa [/bib_ref] Prisons are known to be high-risk settings for the spread of sexually transmitted infections (STIs). A study from Ghana described high prevalence rates of HIV (5.9%), syphilis (16.5%) and Hepatitis B surface antigen (HBsAg) (25.5%) among inmates. [bib_ref] Correlates of HIV, HBV, HCV and syphilis infections among prison inmates and..., Adjei [/bib_ref] Also, a study from Indiana (USA) described higher rates of STIs in a cohort of individuals released from prison than in the general population, with rates in the 1 year after setting free being 2 to 7 times higher for chlamydia, 5 to 24 times higher for gonorrhoea and 19 to 32 times higher for syphilis compared with rates in the general population. Inmates, whether they enter the correctional system with STIs or contract them while in prison, become a risk to public health once they are released. [bib_ref] Test positivity for chlamydia, gonorrhea, and syphilis infection among a cohort of..., Wiehe [/bib_ref] The correctional system should be deeply committed to STIs testing and STIs reduction techniques in prisons, because the STI epidemic continues to be part of reality in several prisons. Historically in Brazil, healthcare for inmates has been performed in fragmented and vertical programmes (routine immunisation, screening for tuberculosis, prevention of HIV and other sexually transmitted infections). There is not an infectious or chronic diseases Open access prevention programme in Brazilian prisons. The country does not have a general picture of the situation and most of the published data in Brazil concerning the incarcerated population have mainly focused on HIV/AIDS, viral hepatitis and tuberculosis and are from specific states or regions. [bib_ref] Active and latent tuberculosis in Brazilian correctional facilities: a cross-sectional study, Carbone [/bib_ref] [bib_ref] A Cross-Sectional Survey of HIV Testing and Prevalence in Twelve Brazilian Correctional..., Sgarbi [/bib_ref] [bib_ref] Prevalence and Incidence of HCV Infection among Prisoners in Central Brazil, Puga [/bib_ref] It is necessary to analyse the situation of this population from a larger perspective/scope, since other health problems are present, such as dengue, with changes in behaviour during epidemic peaks, requiring specific approaches. Planning and executing healthcare programmes for the prison population is a global concern. Prisoners also have high levels of mental illness, chronic disease and drug use, above those of their populations of origin. Health problems resulting from other conditions of confinement have not been the object of healthcare actions, which would allow incarcerated people to have access to healthcare in an integral and effective way. In Brazil, these diseases are not included in the list of mandatory reporting diseases. In addition to these facts, the prisoners' conjugal visits can result in the circulation of pathogens between the prison system and the community surrounding it. [bib_ref] Health disparities and the criminal justice system: an agenda for further research..., Binswanger [/bib_ref] The use of secondary data presents some limitations, that is, the use of information contained in systems such as SINAN, SIM, INFOPEN and GEO Prisons, may not reflect the magnitude of diseases among prisoners. The problem of under-reporting and inadequacy of data completeness is known; however, the information contained in these information systems can provide an initial basis for monitoring and analysing the health situation in a population that has no other data sources available. This information originates in healthcare units and data can take a long time to arrive at the Ministry of Health what can cause delayed notification. The use of multiple sources to retrieve the information of disease cases in the prison systems helps considerably to diminish the problems but does not replace the most accurate picture that could have been available if follow-up information of all exposed individuals was available.
Therefore, although currently available data may be underestimating the real magnitude of health problems in Brazilian prisons, it is the best available and can help for planning appropriate prevention and healthcare strategies addressed to this vulnerable population. Prisoners in Brazil do not have adequate access to healthcare services and neither is properly reached by the healthcare system. The use of SINAN as a monitoring tool is of great relevance for planning healthcare strategies in Brazil, and this applicability must be transferred to the prison system since it will allow reorienting attention to policies concerning the health of incarcerated people.
The increase of diseases notification among prisoners is a signal of the improvement of the healthcare for this population and can contribute for the better quality of care. These data are critical to access the quality of the health system's response and evaluate the vulnerability of the incarcerated population for accessing health care. [bib_ref] Sanitation rights, public law litigation, and inequality: a case study from Brazil, De Barcellos [/bib_ref] The National Policy for Integral Attention to the Health of Incarcerated Persons in the Prison System in foresees the implementation of primary health units in prisons with more than 100 people, and the incorporation of a multiprofessional team composed at least of five professionals (physician, nurse, psychologist, social worker and dentist) and one undergraduate professional (nursing technician). In prisons with up to 100 prisoners, a health team designated by the local Municipal Health Department must perform healthcare. Access to complex levels of healthcare provided for the National Policy requires to be agreed on and defined within each state government.The urgency to implement a public policy of social inclusion focused on promoting human rights for this population indicates the importance of reorienting the healthcare model. The formulation of the National Health Policy in the Penitentiary System in Brazil is an initiative aimed to guarantee the constitutional rights of healthcare and access to the National Health System for prisoners with equity, integrality and universality; additionally, this policy will contribute to organising actions and health services.
Public health must be combined with criminal justice to offer medical care to prison populations, because, generally, they come from areas of society with significant levels of deficient health and social exclusion. Prisoners tend to have poorer physical, mental and social health compared with the general population. [bib_ref] Health disparities and the criminal justice system: an agenda for further research..., Binswanger [/bib_ref] Many prisoners have had little or no regular contact with health services before entering prison. Mental illness, drug dependence and communicable diseases are the prevailing health problems among prisoners. Health caring for prisoners is an essential task and its main activities are critical. However, a full primary care service also includes elements of disease prevention and health promotion and should be accomplished by an interdisciplinary team with sufficient qualified personnel acting in full clinical independence. Measures to implement synchronised information between the criminal justice and public health organisations could generate improvements on the access to healthcare and medical information of prisoners facilitating the transition of care between the healthcare during imprisonment and after prisoners' releasing. [bib_ref] How health care reform can transform the health of criminal justice-involved individuals, Rich [/bib_ref] It is important to ensure that medical files are transferred to healthcare services of the accepting institutions on assignment of a prisoner, as well as subject to medical confidentiality. 1
# Conclusions
Data regarding the health situation in prisons are important for properly developing public policies, determining priorities, guiding the planning and accomplishment of the public health system actions. Physicians and other healthcare professionals working in the prison environment should be included in strategies to reduce health disparities in this population; these professionals should receive continuous training to better understand Open access the peculiarities of healthcare in prisons, the diseases notification system, the prison environment and different types of prison unities (closed, semiopen and open regimes). Prison health services should not be isolated but integrated into the regional and national health and justice systems.
Contributors FJN contributed to the study design and data collection and revised the final version of the manuscript. RBM and CPG contributed to the literature review and database revision and revised the final version of the manuscript. EZ and RdAC conducted the data analysis, contributed to the results section and revised the final version of the manuscript. AEM designed the study, supervised the design of the questionnaire plan and data collection and the writing and revised the final version of the manuscript. All authors read and approved the final manuscript.
Funding Brazilian Ministry of Health provided funding support for this study: Cooperation term #181/2013, Process #25000.184518/2013-41.
Competing interests None declared.
Patient consent for publication Not required.
ethics approval This study was submitted to the Ethics Committee of the Health Sciences Center in the Federal University of Espirito Santo and it was approved under number 1.058.616/2015.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Database used in this paper belongs to the Brazilian Ministry of Health. Data can be available upon request after contacting Ronaldo Coelho at ronaldo. coelho@ aids. gov. br.
Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http:// creativecommons. org/ licenses/ by-nc/ 4. 0/.
## References
[fig] Figure 1: Historical series of the proportion of prisons' population by the annual population per 1000 people, Brazil, 2007-2014. [/fig]
[fig] Figure 2: Trends of compulsory notification diseases reported among prisons' population in Brazil, 2007-2014. [/fig]
[table] Table 1: Linear regression model with the proportion of cases reported by geographical regions and states ofBrazil, 2007Brazil, - 2014 One state from the North Region and three states from the Northeast region did not notify any case of disease during the period. [/table]
[table] Table 3: Linear regression model of reported diseases proportion and its trends, 2007-2014 [/table]
|
Human Bocavirus in Children Hospitalized for Acute Gastroenteritis: A Case-Control Study
Background. Human bocavirus (HBoV) was recently discovered in children with respiratory tract disease and gastroenteritis. The causative role of HBoV in human gastroenteritis remains uncertain, and, to our knowledge, no previous case-control study has studied the relationship between HBoV and gastroenteritis.Methods. We conducted a case-control study that examined stool samples from 397 children with diarrhea and from 115 asymptomatic control subjects. HBoV was detected using polymerase chain reaction. Real-time polymerase chain reaction was used to quantify the HBoV loads in case and control groups. Common enteric viruses were examined using enzyme-linked immunosorbent assays, polymerase chain reaction, and reverse-transcription polymerase chain reaction.Results. At least 1 viral agent was discovered in 60.2% of cases. HBoV was detected in 14 samples, and 9 were coinfected with either rotavirus (7 of 14 samples) or human calicivirus (2 of 14). Many (8 [57.1%] of 14) of the HBoV infections occurred during September-December 2006. Most (12 [85.7%]) of the HBoV-infected children were 7-18 months of age. The percentage of children with HBoV infection did not differ significantly between case patients and control subjects (3.5% vs. 3.5%), and the statistical analysis did not support a correlation between HBoV infection and more-severe clinical symptoms. The viral load differences between the 2 groups were not statistically significant ( , by log-normal Student's t test). In addition, the VP1/VP2 partial gene of HBoV P p .09 from case patients and control subjects showed minimal sequence variation.Conclusions.A single genetic lineage of HBoV was revealed in persons in China. Despite its high prevalence in stool samples, our study does not support a causative role of HBoV in gastroenteritis.
Gastroenteritis is a major cause of childhood morbidity and mortality worldwide. Between 1.5 and 2 million infants and young children die of gastroenteritis-related diseases or complications annually. Pathogenic enteric bacteria are important etiologic agents of this disease, but with the improvement of molecular detection methods, a large proportion of cases of gastroenteritis is found to be associated with viral etiological agents [bib_ref] Identification of viral agents causing diarrhea among children in the Eastern Center..., Fodha [/bib_ref] [bib_ref] Diversity of viruses associated with acute gastroenteritis in children hospitalized with diarrhea..., Nguyen [/bib_ref] , such as rotaviruses, human calicivirus (HuCV), adenoviruses, and astroviruses [bib_ref] Viruses causing gastroenteritis, Wilhelmi [/bib_ref]. However, the etio-logic agents are still not identified in some patients with gastroenteritis, despite improvements in diagnostic technology.
In 2005, Allander et al. [bib_ref] Cloning of a human parvovirus by molecular screening of respiratory tract samples, Allander [/bib_ref] reported the detection of a new human parvovirus in children with acute respiratory tract infections. This virus belongs to the genus Bocavirus in the subfamily parvovirinae of the family parvoviridae and is most closely related to bovine parvovirus and minute virus of canines. Therefore, it was named "human bocavirus" (HBoV). Subsequently, HBoV has been detected frequently in children with respiratory tract infections and asthma exacerbation worldwide [bib_ref] Detection of viruses identified recently in children with acute wheezing, Chung [/bib_ref] [bib_ref] Human bocavirus in French children, Foulongne [/bib_ref] [bib_ref] Human bocavirus infection among children, Kaplan [/bib_ref] [bib_ref] Detection of human bocavirus in Japanese children with lower respiratory tract infections, Ma [/bib_ref] [bib_ref] Human bocavirus infection, People's Republic of China, Qu [/bib_ref] [bib_ref] Evidence of human coronavirus HKU1 and human bocavirus in Australian children, Sloots [/bib_ref]. Several studies confirmed that HBoV infection is indeed associated with respiratory tract symptoms [bib_ref] Human bocavirus and acute wheezing in children, Allander [/bib_ref] [bib_ref] Human bocavirus infection in young children in the United States: molecular epidemiological..., Kesebir [/bib_ref] [bib_ref] Epidemiological profile and clinical associations of human bocavirus and other human parvoviruses, Manning [/bib_ref] [bib_ref] Human bocavirus: a novel parvovirus epidemiologically associated with pneumonia requiring hospitalization in..., Fry [/bib_ref]. Recently, HBoV has also been implicated in diarrhea, and its detection rates in children with gastroenteritis have a range of 0.8%-9.1%. Therefore, this newly identified virus has been suggested to be an enteric pathogen, as well as being a respiratory pathogen [bib_ref] Human bocavirus: prevalence and clinical spectrum at a children's hospital, Arnold [/bib_ref] [bib_ref] Clinical and molecular epidemiology of human bocavirus in respiratory and fecal samples..., Lau [/bib_ref] [bib_ref] Human bocavirus, a respiratory and enteric virus, Vicente [/bib_ref] [bib_ref] Detection of human bocavirus in children hospitalized because of acute gastroenteritis, Lee [/bib_ref]. However, because most studies assessed the prevalence of HBoV in children with gastroenteritis in the absence of defined control children without enteric disease, the clinical spectrum of HBoV and the role that it plays in gastroenteritis remain to be clarified. The present case-control study investigated the prevalence of HBoV and other major viral etiologic agents in children hospitalized for gastroenteritis from July 2006 through September 2007 in Lanzhou, China, and the relationship between HBoV and gastroenteritis.
## Materials, patients, and methods
## Study participants and sample collection and processing.
From July 2006 through September 2007, we conducted a casecontrol study to explore gastroenteritis-associated viral agents in hospitalized children with diarrhea and in healthy children in Lanzhou, China. Stools from 397 hospitalized children with diarrhea and from 115 asymptomatic control subjects were collected. All of the children were !5 years of age, and their parents were interviewed to determine the symptoms. Consecutive cases were collected from among children who were hospitalized for gastroenteritis in the Department of Pediatrics, The First Hospital of Lanzhou University (Lanzhou, China). Diarrhea was defined as у3 loose stools in the previous 24-72 h [bib_ref] Incidence of Norwalk-like viruses, rotavirus and adenovirus infection in patients with acute..., Subekti [/bib_ref]. Patients were excluded from the study if insufficient stool samples were available for a complete evaluation of viral agents, their stools had clear blood streaks, or they had another diagnosed illness, such as pneumonia. Control subjects were asymptomatic children who visited the First Hospital of Lan- zhou University Pediatric Primary Care Center for a routine examination and did not have fever, diarrhea, vomiting, or a respiratory illness in the previous 3-week period [bib_ref] Detection by PCR of eight groups of enteric pathogens in 4,627 faecal..., Amar [/bib_ref]. All of the control subjects in this study were selected by the frequencymatching method. Control subjects received follow-up by telephone, and the children who had had the above-mentioned clinical symptoms during the week after the initial examination were excluded. Informed consent was obtained from the parents of all of the children who provided specimens. The study protocol was approved by the hospital ethics committee. All specimens were stored at Ϫ70ЊC until further analysis. Viral RNA and DNA were extracted from 140 mL of 10% fecal suspension in phosphate-buffered saline with use of the QIAamp Viral RNA Mini Kit (Qiagen), which is supposed to extract viral RNA and DNA simultaneously, according to the manufacturer's instructions.
Rotavirus detection. Rotavirus antigen was detected using a commercial ELISA kit (IDEIA Rotavirus; Dako), according to the manufacturer's instructions. Positive specimens were then G and P genotyped with use of nested PCR with typespecific primers, as described elsewhere [bib_ref] Polymerase chain reaction amplification and typing of rotavirus nucleic acid from stool..., Gouvea [/bib_ref].
Detection of common enteric viruses. Six enteric viruses were identified using multiplex RT-PCR and PCR with 3 sets of primers [bib_ref] Development of a novel protocol for RT-multiplex PCR to detect diarrheal viruses..., Phan [/bib_ref] , as follows: set A, for detecting noroviruses GI and GII, sapoviruses, and astroviruses; set B, for all adenoviruses; and set C, for group B and C rotaviruses. The multiplex RT-PCR and PCR were performed in accordance with a protocol described elsewhere [bib_ref] Development of a novel protocol for RT-multiplex PCR to detect diarrheal viruses..., Phan [/bib_ref].
## Hbov detection.
HBoV was detected in the extracted DNA by PCR amplification of a 291-base pair fragment of the NS1 gene with forward primer HBOV-1 (5 -TATGGCCAAGGCAA-TCGTCCAAG-3 ) and reverse primer HBOV-2 (5 -GCCGCGT-GAACATGAGAAACAGA-3 ), as described elsewhere [bib_ref] Evidence of human coronavirus HKU1 and human bocavirus in Australian children, Sloots [/bib_ref]. To acquire the partial VP1/VP2 gene, we used forward primer VP1/ 2-1 (5 -GGACCACAGTCATCAGAC-3 ) and reverse primer VP1/2-2 (5 -CCACTACCATCGGGCTG-3 ), targeting an 820base pair fragment [bib_ref] Human bocavirus infection in young children in the United States: molecular epidemiological..., Kesebir [/bib_ref]. The reaction mix contained 20 pmol of each primer and 2.5 units of ExTaq DNA polymerase (Takara Bio). After 5 min at 94ЊC, 35 cycles of amplification (94ЊC for 1 min, 54ЊC for 1 min, and 72ЊC for 2 min) were performed, followed by a 10-min extension at 72ЊC. Positive PCR products were purified using a QIAquick PCR purification kit (Qiagen) and were sequenced by Invitrogen.
Sequence analysis and accession numbers. The nucleotide and deduced amino acid sequences of the NS1 gene and the VP1/VP2 partial gene were compared with those of HBoV strains available at the GenBank site. Phylogenetic analyses were conducted with MEGA, version 3.1. The 12 partial sequences of the VP1/VP2 gene were submitted toGenBank (accession numbers EU400116-EU400128). and 2.5 mL of sample DNA. PCR was conducted at 50ЊC for 2 min and at 95ЊC for 10 min, followed by 50 cycles at 95ЊC for 15 s and at 60ЊC for 1 min. Serial dilutions of the pGEM-T Easy vector (Promega) containing the HBoV NS1 gene were used as a quantification standard. Positive and negative controls were included. The Rotor-Gene 3000 real-time thermal cycling system (Corbett Research) was used, and the data were analyzed using Rotor-Gene software (Corbett Research), with the help of a standard curve. The minimum viral load that would allow reproducible quantification was 10 copies per reaction.
Statistical analysis. The statistical significance of rates between various groups was tested using the x 2 test and Fisher's exact test. The statistical significance of means between various groups was tested using the Student's t test and the log-normal Student's t test. Analyses were performed using SPSS, version 11.5 (SPSS).
# Results
## Patient characteristics.
The ages of children with diarrhea ranged from 15 days to 60 months (mean age ע SD, months). The majority of patients (74.8%) were 11.18 ע 9.3 0-12 months of age. The ratio of boys to girls was 1.8:1.
Virological findings in children with diarrhea. Of the 397 specimens, 239 (60.2%) contained at least 1 viral agent. Rotaviruses were identified in 165 (41.6%) of the 397 samples: G1 was the most common G serotype (91 of 165; 55.2%), and P was the most common P genotype (112 of 165; 67.9%) [bib_ref] Human bocavirus infection among children, Kaplan [/bib_ref].
We detected HuCV in 52 (13.1%) of 397 samples, astroviruses in 13 samples (3.3%), adenoviruses in 22 samples (5.5%), and HBoV in 14 samples (3.5%). Group B and C rotaviruses were not detected in this study. Detection of rotaviruses, HuCV, astroviruses, and HBoV peaked during the winter. HBoV was not detected during spring 2007. Of the adenovirus-positive samples, 77.3% (17 of 22 samples) were collected during July-September 2007, indicating a possible outbreak of adenovirus infection during that season (figure 1). The majority of patients who were positive for rotaviruses, astroviruses, and adenoviruses were 0-12 months of age, whereas HuCV and HBoV were found mostly in children aged 7-18 months (figure 2). Of the 14 HBoV-positive samples, 64.3% (9) were coinfected with either rotavirus [bib_ref] Human bocavirus in French children, Foulongne [/bib_ref] or HuCV [bib_ref] Identification of viral agents causing diarrhea among children in the Eastern Center..., Fodha [/bib_ref]. Of the patients with HBoVpositive samples, 9 were boys and 5 were girls. Of these 14 patients, 7 had fever and 11 were vomiting, and the mean duration and frequency of diarrhea were 4.6 days and 6.5 times per day, respectively [fig_ref] Table 1: Clinical characteristics and viral loads of the 18 case and control children... [/fig_ref].
Prevalence of HBoV in asymptomatic children. The ages of 115 asymptomatic control children ranged from 1 to 60 months (mean age ע SD, months). The majority 14.1 ע 17.2 of these children (78.3%) were 1-18 months of age. The ratio of boys to girls was 1.2:1. The differences in mean age (P p , by Student's t-test) and sex ( , by x 2 test) between .08 P p .07 the case group and the control group were not statistically significant. HBoV was detected in 4 (3.5%) of 115 samples derived from the control subjects; 4 HBoV-positive cases were confirmed in subjects aged р6 months. Samples were collected in March, May, June, and August. The ratio of boys to girls was 1:1.
Phylogenetic analyses of HBoV. All PCR products of the NS1 gene were confirmed by sequencing. Because a previous study demonstrated that the greatest variation in the HBoV genome was in the VP1/VP2 gene, especially at its 3 end [bib_ref] Human bocavirus infection in young children in the United States: molecular epidemiological..., Kesebir [/bib_ref] , 12 HBoV-positive specimens were randomly selected to amplify VP1/VP2 partial gene sequences (nucleotides 4370-5189 of the HBoV genome). Phylogenetic analyses indicated that all 12 VP1/VP2 genes in our study were in the same cluster as other strains from China (with 199% sequence homology) [fig_ref] Figure 3: Phylogenetic tree of the VP1/VP2 gene sequences of 12 human bocavirus strains... [/fig_ref].
Quantitative analysis of HBoV DNA. We used real-time PCR to quantify the HBoV load. The mean HBoV load in the case group was copies/mL of extract (range, [bib_ref] Detection of viruses identified recently in children with acute wheezing, Chung [/bib_ref] 2.0 ϫ 10 -copies/mL), and the mean HBoV load in 2.6 ϫ 10 1.2 ϫ 10 the control group was copies/mL of extract (range, [bib_ref] Human bocavirus in French children, Foulongne [/bib_ref] 1.5 ϫ 10 -copies/mL). There was no statitistically 1.8 ϫ 10 1.3 ϫ 10 signficant difference between the mean values of these 2 groups ( , log-normal Student's t test) (tables 1 and 2). P p .09
Association of HBoV with gastroenteritis. In our study, none of the HBoV-positive patients had respiratory symptoms. HBoV was not more prevalent among children with gastroenteritis than among asymptomatic children (3.5% vs. 3.5%) (table 2). HBoV was not more prevalent among children with gastroenteritis of undetectable etiology than among those in whom other viruses were detected (3.1% vs. 3.8%;
, by P p .75 x 2 test) [fig_ref] Table 3: Comparison of detection rates between children with gastroenteritis of undetectable etiology and... [/fig_ref]. When the HBoV-positive and HBoV-negative case groups were compared, neither the rates of fever among patients nor the rates of vomiting among patients differed statistically significantly (
, by x 2 test; and , by P p .80 P p .06 Fisher's exact test; respectively). In addition, the mean duration and frequency of diarrhea did not differ statistically significantly ( and , respectively, by Student's t test). Because P p .52 P p .42 the seasonal and age distributions of HBoV were similar to those of rotavirus and because there was often coinfection with HBoV and rotavirus, we compared the clinical symptoms be-tween the group of children infected with rotavirus alone and the group of children coinfected with rotavirus and HBoV, and we did not find a statistically significant difference between these 2 groups (all , by x 2 test and Student's t test) (table P 1 .05 4). These results suggest that infection with HBoV did not exacerbate the clinical symptoms of gastroenteritis.
# Discussion
This study examined common viral etiologic agents, to delineate the clinical role played by HBoV in children hospitalized with gastroenteritis. Our results revealed that enteric viruses play an important role in pediatric diarrhea and that the most common viral agents causing severe gastroenteritis consisted of group A rotaviruses, followed by HuCV. The prevalence of adenoviruses and astroviruses was similar to that reported elsewhere [bib_ref] Diversity of viruses associated with acute gastroenteritis in children hospitalized with diarrhea..., Nguyen [/bib_ref] [bib_ref] Influence of enteric viruses on gastroenteritis in Albania: epidemiological and molecular analysis, Fabiana [/bib_ref] [bib_ref] Viral gastroenteritis in children hospitalised in Sicily, Italy, Colomba [/bib_ref] [bib_ref] Molecular epidemiology and clinical manifestations of viral gastroenteritis in hospitalized pediatric patients..., Chen [/bib_ref]. Although some studies have detected high rates of groups B and C rotaviruses in stool samples from children with diarrhea [bib_ref] Diversity of viruses associated with acute gastroenteritis in children hospitalized with diarrhea..., Nguyen [/bib_ref] [bib_ref] RT-PCR based diagnosis revealed importance of human group B rotavirus infection in..., Barman [/bib_ref] [bib_ref] Virus diversity and an outbreak of group C rotavirus among infants and..., Phan [/bib_ref] , these pathogens were not found in samples from our participants.
Previous findings [bib_ref] Minute virus of canines (MVC, canine parvovirus type-1): pathogenicity for pups and..., Carmichael [/bib_ref] [bib_ref] Pathological and virological studies of experimental parvoviral enteritis in calves, Durham [/bib_ref] [bib_ref] An update on aspects of viral gastrointestinal diseases of dogs and cats, Squires [/bib_ref] supported the association of animal bocaviruses with both respiratory symptoms and gastroenteritis, particularly in calves and puppies. Some studies raised the question about whether HBoV is a cause of human gastroenteritis [bib_ref] Human bocavirus: prevalence and clinical spectrum at a children's hospital, Arnold [/bib_ref] [bib_ref] Real-time PCR for diagnosis of human bocavirus infections and phylogenetic analysis, Neske [/bib_ref]. HBoV is thought to be swallowed during a respiratory tract infection and subsequently excreted in the feces, without further replication in the gastrointestinal tract [bib_ref] Real-time PCR for diagnosis of human bocavirus infections and phylogenetic analysis, Neske [/bib_ref]. Recently, HBoV was found to occur frequently in stool samples from children with gastroenteritis. Vicente et al. [bib_ref] Human bocavirus, a respiratory and enteric virus, Vicente [/bib_ref] found that 48 (9.1%) of 527 stool samples from Spanish patients with gastroenteritis were positive for HBoV. Lau et al. [bib_ref] Clinical and molecular epidemiology of human bocavirus in respiratory and fecal samples..., Lau [/bib_ref] detected HBoV in 2.1% of fecal samples from children with gastroenteritis in Hong Kong, and some of these patients had respiratory tract symptoms. These studies implicated HBoV as a gastroenteritis-associated enteric virus. However, the con-current detection of HBoV and other enteric viruses, which is unusual for known pathogens, raises concern over a causative role of HBoV in human gastroenteritis. Therefore, additional evidence is required to establish a link between HBoV and gastroenteritis.
In our study, the detection rate of HBoV (3.5%) was comparable to previously reported rates [bib_ref] Clinical and molecular epidemiology of human bocavirus in respiratory and fecal samples..., Lau [/bib_ref] [bib_ref] Human bocavirus, a respiratory and enteric virus, Vicente [/bib_ref] [bib_ref] Detection of human bocavirus in children hospitalized because of acute gastroenteritis, Lee [/bib_ref]. We also found that HBoV infections occurred more frequently during the winter months (50% of cases) in children aged !2 years. Coinfection with HBoV and another virus was common (64.3% among all HBoV-positive children with gastroenteritis). The most frequent codetection of HBoV and rotavirus further supported the idea that HBoV is an "innocent bystander" in gastroenteritis. In this regard, our results were generally consistent with previous findings [bib_ref] Evidence of human coronavirus HKU1 and human bocavirus in Australian children, Sloots [/bib_ref] [bib_ref] Human bocavirus, a respiratory and enteric virus, Vicente [/bib_ref] [bib_ref] Human bocavirus in hospitalized children, Smuts [/bib_ref] , and the similar rates of HBoV in our control (3.5%) and gastroenteritis (3.5%) groups also argue against a causative role of HBoV in gastroenteritis. Furthermore, the difference in the HBoV loads, as measured in our control and gastroenteritis groups, was statistically nonsignificant.
On the basis of VP1/VP2 gene sequences, all HBoV isolates found in our study were in the same cluster as the original isolate ST2 identified by Allander et al. [bib_ref] Cloning of a human parvovirus by molecular screening of respiratory tract samples, Allander [/bib_ref] , with 199% DNA sequence homology. Moreover, the other HBoV strains identified in China-including CZ, WL, BJ, and HK-are in the same cluster. This suggests that a single genetic lineage of HBoV is circulating in humans in China and that the HBoV viruses found in both the respiratory and enteric tracts of humans in China probably belong to a single genetic lineage.
To our knowledge, this is the first case-control study to investigate the linkage between HBoV and gastroenteritis. In summary, we did not find statistically significant differences in the prevalence and HBoV loads between children with and without gastroenteritis. Concurrent detection of HBoV and other enteric pathogens, such as rotaviruses, in children with gastroenteritis was common. In addition, infection with HBoV did not significantly influence the severity of gastroenteritis. Our results do not support a causative role for HBoV in gastroenteritis, despite its frequent detection in fecal samples worldwide. To further clarify the linkage between HBoV and gastroenteritis, future priorities should include additional case-control studies that have larger cohorts of case and control samples, the establishment of serological methods to trace the viral antigens and the immune response, and the development of an animal model for HBoV infection.
[fig] Figure 1: Seasonal distribution of viral gastroenteritis in pediatric patients hospitalized in the First Hospital of Lanzhou University (Lanzhou, China). HBoV, human bocavirus; HuCV, human calicivirus. [/fig]
[fig] Figure 2: Age distribution of pediatric patients with viral gastroenteritis who were hospitalized in the First Hospital of Lanzhou University (Lanzhou, China). HBoV, human bocavirus; HuCV, human calicivirus. [/fig]
[fig] Figure 3: Phylogenetic tree of the VP1/VP2 gene sequences of 12 human bocavirus strains from fecal samples. Strains 53040-LanZhou, 53047-LanZhou, 53055-LanZhou, and 53889-LanZhou were from asymptomatic children; the other strains were from children with gastroenteritis. Human bocavirus strains ST1 and ST2 from Sweden are the 2 prototype strains. Strains HK1, BJ3722, WLL-1, and CZ643 are from China, and strain CRD2 is from the United States. [/fig]
[table] Table 1: Clinical characteristics and viral loads of the 18 case and control children who were positive for human bocavirus. NOTE. Duration of diarrhea indicates the time from the first day with diarrhea to the day of hospitalization. +, Present; Ϫ, absent. [/table]
[table] Table 2: Comparison of detection rates and human bocavirus (HBoV) loads between case group and control group. [/table]
[table] Table 3: Comparison of detection rates between children with gastroenteritis of undetectable etiology and those in whom other viruses were detected.Table 4. Comparison of clinical symptoms between different groups of children with gastroenteritis. [/table]
|
Crystal structure of the DNA-bound VapBC2 antitoxin/toxin pair from Rickettsia felis
Supplementary Figure 1 a b Supplementary Figure 3 Chain X Chain Y K19(G)* NZ NZ R22(G) NH1 NH1 Q11(F) N N
Supplementary Figure 1 a b Supplementary Figure 3 Chain X Chain Y K19(G)* NZ NZ R22(G) NH1 NH1 Q11(F) N N
A T G C T A A T T A A T T A A T T A T A A T A T 5'
Stereo views of the superposition of the VapB2 homodimer (gold) to other antitoxins of the swappedhairpin β-barrel family: a) the transition state regulatory protein AbrB (green, PDB ID 2ro4) and b) the antitoxin MazE (light blue, PDB ID 1ub4). The β-barrel core of AbrB and MazE were superposed to the VapB2 homodimer made by monomers E and H. The DNA bound to VapB2 is shown in pink as reference. Proteins are depicted as α-carbon trace and DNA as ribbons. The two views are mutually perpendicular.SupplementaryFigure 2 a b Analysis of the DNA conformation in the VapBC2-DNA complex. a) Size of the major (black triangles and continuous line) and the minor (empty circles and dashed line) grooves. b) Roll angles. Calculations were performed using the program Curves+ 42 . |
Construction of Classifier Based on MPCA and QSA and Its Application on Classification of Pancreatic Diseases
A novel method is proposed to establish the classifier which can classify the pancreatic images into normal or abnormal. Firstly, the brightness feature is used to construct high-order tensors, then using multilinear principal component analysis (MPCA) extracts the eigentensors, and finally, the classifier is constructed based on support vector machine (SVM) and the classifier parameters are optimized with quantum simulated annealing algorithm (QSA). In order to verify the effectiveness of the proposed algorithm, the normal SVM method has been chosen as comparing algorithm. The experimental results show that the proposed method can effectively extract the eigenfeatures and improve the classification accuracy of pancreatic images.
# Introduction
Pancreatic carcinoma is a frequent digestive tract tumor. The malignant degree of this kind of cancer is always very high, and it is difficult to be early diagnosed and treated. Due to the fact that pancreatic carcinoma is often diagnosed when it is advanced, very few pancreatic tumors can be removed by operation. As we know, many famous people died of this disease. So, it is necessary to diagnose pancreatic carcinoma as early as possible. Computer-aided diagnosis (CAD) [bib_ref] Computer-aided diagnosis: the emerging of three CAD systems induced by Japanese health..., Fujita [/bib_ref] technology was established with the development of imageprocessing technology and pattern recognition technology. Researching of CAD technology shows that CAD can provide advisory opinions for the doctor and help to improve the diagnostic rate. With the development of medical imaging, it is important to represent the pancreas by a model and it is also important to try to distinguish different appearance of pancreas.
Tensors are geometrical quantity that is used to describe linear relations among vectors, scalars, and other tensors. In this paper, the pancreas CT images can be treated as several third-order tensors, and then we extract the feature to gain the eigentensors for classification.
Principal component analysis (PCA) [bib_ref] Face recognition using Eigenfaces, Kshirsagar [/bib_ref] is a famous method used in the recognition of subspace, which is one of the classical methods based on statistical feature. The core idea of PCA is to reduce the dimensionality of a dataset that consisted of a larger number of interrelated variables and, in the meantime, try to retain the variation in the original dataset as much as possible. But this method has two problems.
(1) Basic using of PCA to transfer tensor objects to highdimension vector (vectorization) obviously results in high cost of processing and memory in next step [bib_ref] MPCA: multilinear principal component analysis of tensor objects, Lu [/bib_ref]. For example, if there is a gray image (640 × 640), the vector which the image transfers to will be 409600 × 1.
(2) In using PCA, reshaping breaks the natural structure and correlation in the original data [bib_ref] MPCA: multilinear principal component analysis of tensor objects, Lu [/bib_ref] , which may affect the subsequent operation and lead to bad results.
In order to solve these problems, this paper uses multilinear principal component analysis (MPCA) referred to in [bib_ref] MPCA: multilinear principal component analysis of tensor objects, Lu [/bib_ref]. MPCA follows the classical PCA paradigm and multilinear algorithm, to ensure the fact that it is able to reduce all the tensor dimensionality and that it is also able to get more variational forms among the original tensors with tensorial mapping to keep original structure and correlation [bib_ref] MPCA: multilinear principal component analysis of tensor objects, Lu [/bib_ref]. Support vector machine (SVM) [bib_ref] Nesting support vector machinte for muti-classification, Liu [/bib_ref] is commonly used to train a classifier. The main factor to affect the classification performance is the parameters used in SVM. Recently, there are many algorithms for SVM parameters optimization, such as ant colony (ACO) algorithm [bib_ref] Ant system: optimization by a colony of cooperating agents, Dorigo [/bib_ref] , simulated annealing (SA) algorithm, genetic algorithm (GA) [bib_ref] Quantum-inspired genetic algorithms, Narayanan [/bib_ref] , and quantum genetic algorithm (QGA) [bib_ref] An improved real coded quantum genetic algorithm and its applications, Xu [/bib_ref] [bib_ref] A variant of quantum genetic algorithm and its possible applications, Tiwari [/bib_ref]. SA is a generic probabilistic algorithm, which is good at locating the optimal solution of the variable in a large search space. The advantages of SA are described as follows.
(1) The objective function can be nonlinear, discontinuous, and random. (2) The objective function can have any boundary conditions and constraints. (3) The programming workload of SA is low, so that it is easy to be implemented. (4) In statistics, we can find the optimal solutions. But there are also some problems of SA. For example, rapid cooling can lead to simulation hardening which cannot be ensured to find the optimal solution. Quantum evolution algorithm (QEA) [bib_ref] Quantum-inspired evolutionary algorithm for a class of combinatorial optimization, Han [/bib_ref] is also a probability optimization algorithm. QEA has good searching ability for low dimensional function. However, it is not good for high dimensional functions. Therefore, in this paper, we use quantum simulated annealing algorithm (QSA) [bib_ref] Quantum annealing algorithms: atate of the art, Du [/bib_ref] , which is the combination of SA and QEA, to optimize SVM parameters for training classifier of pancreatic diseases. This paper is organized as follows. Section 2 introduces the proposed method; firstly, we will explain the construction of high-order tensor, then we briefly introduce the method of MPCA for feature extraction, and finally we introduce the support vector machine of quantum simulated annealing algorithm optimization (QSA-SVM) for classification. Section 3 presents the construction of pancreas images after MPCA and the results of classification, and we discuss the future of clinical implications of the results. In Section 4, we conclude the works in this paper.
# Materials and method
In this section, firstly we will introduce the whole procedure of the proposed method, which is shown in [fig_ref] Figure 1: The main flow of the proposed method [/fig_ref] , and then we give a detailed explanation of each process.
## 2-mode (row)
3-mode (thickness) 1-mode (column) : Illustration of the pancreatic CT image as a third-order tensor.
The process of the proposed method is as follows.
(1) Image preprocessing: first, we segment the CT images of abdomen to gain the pancreas region of image, and then we normalized the images after segmentation.
(2) High-order tensors construction: at first, we collect a group of pancreatic images and then combine them into a new dataset.
(3) The feature extraction: in this paper, we use the method of MPCA to extract the eigentensors for classification.
(4) Pancreas diseases classification based on QSA-SVM: after we obtain the eigentensors by MPCA, we can treat the eigentensors as samples, and then we use the approach of SVM optimized by QSA to classify pancreas diseases.
## Construction of tensors.
We treat the segmented pancreatic CT images as several third-order tensors with the column, row, and thickness modes. In this paper, we treat each CT image as one data sample. Hence, the input is several third-order tensors and the spatial column space, row space, and the thickness space were regarded as its three modes, as shown in . Local optimization The size of each image is standard 128 × 128; the thickness of the CT image is 2.77 mm. Before providing the samples to MPCA, the tonsorial inputs need to be normalized to the same dimension in each mode, so the three modes of the tensor are normalized by default, and we can consider one sample as ∈ 1 × 2 × 3 .
[formula] (i) Calculate { = × 1 (1) × 2 (2) ⋅ ⋅ ⋅ × ( ) , = 1, . . . , }. (ii) Calculate Ψ 0 = ∑ =1 2 (iii) [/formula]
## Feature extraction based on mpca.
In this paper, an MPCA [bib_ref] MPCA: multilinear principal component analysis of tensor objects, Lu [/bib_ref] solution to the problem of dimensionality reduction for tensor objects is introduced; its research and analysis are also included. First we provided a series of zero-mean value N-order tensor ∈ 1 × 2 ×⋅⋅⋅× , we need to gain a group of new N-order tensor, ∈ 1 × 2 ×⋅⋅⋅× ( < ), that needs to be closed to the original tensor as much as possible. The procedure of MPCA algorithm is shown in [fig_ref] Figure 3: The flow of MPCA algorithm [/fig_ref].
In the preprocessing phase, we center the input original tensors as = − , = 1, . . . , , where M is the number of the samples, and the tensor mean can be described as follows:
[formula] = 1 ∑ =1 .(1) [/formula]
In the initialization phase, we calculate the eigendecomposition of ( ) * = ∑ =1 ( ) ⋅ ( ) and set the nth mapping matrix ( ) which consists of the eigenvectors corresponding to the most significant eigenvalues, for = 1, . . . , ( = 3).
In the local optimization phase, we will focus on doing the local optimization to obtain the new N-order tensors ; the detailed method of the optimization is given in [bib_ref] MPCA: multilinear principal component analysis of tensor objects, Lu [/bib_ref] and the pseudocode of the method is shown as Pseudocode 1. In the pseudocode, Ψ is the total tensor scatter of , = 1, . . . , , M is the number of the samples, and ( ) can be defined as [bib_ref] Face recognition using Eigenfaces, Kshirsagar [/bib_ref]. In [bib_ref] Face recognition using Eigenfaces, Kshirsagar [/bib_ref] , ( ) is the mean mapping matrix of ( −1) :
[formula] ( ) = ∑ =1 ( ) ⋅ ( ) ⋅ ( ) ⋅ ( ) , ( ) = ( ) − ( ) , ( ) = ( ( +1) ⊗ ⋅ ⋅ ⋅ ⊗ ( + ) ⊗ (1) ⊗ ⋅ ⋅ ⋅ ⊗ ( −1) ) .(2) [/formula]
In the projection phase, we project the centralized eigentensors using the nth mapping matrix ( ) obtained by the local optimization phase to get the new eigentensors . It is shown as follows:
[formula] = × 1 (1) × 2 (2) ⋅ ⋅ ⋅ × ( ) , = 1, . . . , .(3) [/formula]
We used the eigenvector * projected by the eigentensors for classification [bib_ref] MPCA: multilinear principal component analysis of tensor objects, Lu [/bib_ref].
## Construction of the classifier
## Concept of quantum bit and quantum
Gate. The term quantum comes from quantum mechanics. Quantum, which is the general name of all microscopic particles in the microscopic world, is different from the macroscopic object. Its movements obey the statistical law, not the deterministic law. Compared with the classical computing using 0 and 1 to represent information, the quantum computing uses |0⟩, |1⟩ and their superposition state to represent information. The superposition state is as follows:
[formula] ⟩ = |0 ⟩ + |1 ⟩ , s.t. 2 + 2 = 1.(4) [/formula]
The measurement of quantum state can cause the collapse of quantum state, so that the final state can be confirmed. The relationship of quantum state, superposition state, and the collapse caused by measurement is shown in [fig_ref] Figure 4: The relationship of the three states [/fig_ref].
In the quantum computing, the quantum state changes when we have a series of unitary transformations on it. The equipment (a unitary matrix) is called quantum gate which is as follows: We exchange two probability amplitudes of a quantum bit by the quantum gate as follows
[formula] [ cos − sin sin cos ] .(5)[ 0 1 1 0 ] [ ] = [ ] .(6) [/formula]
## Construction of the classifier based on qsa.
We use SVM to train the classifier. SVM can be used to solve some problems, such as the small number of samples, nonlinear, high dimension pattern recognition, and local minimum point, but if the selection of the kernel function parameters, penalty factor , or other parameters is not reasonable, the SVM prediction accuracy will be greatly reduced in classification process. In this paper, QSA is used for optimizing the SVM parameters, penalty factor C, and the parameter of RBF .
We assume that there are m chromosomes in the population and n quantum bits in a chromosome. In QSA, the two probability amplitudes of the quantum bit are treated as the chromosome gene. In the fixed population scale, it can make the search space double, so that the convergence speed will be fast. In fact, the optimal solution is embodied in the optimal probability amplitude of the quantum bit of the optimal chromosome. We assume that the optimal probability amplitude is (cos 1 , cos 2 , . . . , cos ), when another chromosome has the quantum bits ( /2 − 1 , /2 − 2 , . . . , /2 − ), the sine item of this chromosome also has the optimal solution.
In the following description, we set that Φ is the ith chromosome in the population, is the phase, 1 ≤ ≤ , 1 ≤ ≤ . For C, its value ranges from 2 −10 (min c) to 2 9 (max c) and for , its value ranges from 2 −10 (min g) to 2 10 (max g).
The main flow of QSA-SVM is shown in [fig_ref] Figure 5: The flow of QSA-SVM. [/fig_ref].
Step 1. Initialization of parameters.
Step 2. Coding the chromosome using phase, = [ 1 , 2 , . . . , 3 ].
Step 3. Solution space transformation for chromosomes and computing fitness. For the quantum bit, [cos , sin ] , we use the linear transformation [bib_ref] A novel quantum ant colony optimizing algorithm, Yang [/bib_ref] as [bib_ref] Quantum-inspired genetic algorithms, Narayanan [/bib_ref] to transform to the solution space. In (7), is one gene on chromosome, min and max, respectively, are minimum and maximum of in practice, is the corresponding cos solution of phase , and is the sin solution. In our method, can represent the penalty factor C or the parameter of RBF :
[formula] [ ] = 1 2 [ 1 + cos 1 − cos 1 + sin 1 − sin ] [ max min ] , ∈ [min, max] .(7) [/formula]
We use the SVM prediction accuracy as the fitness of chromosomes and leave one out (LOO) to evaluate. Then, we keep all information of the optimal individual.
Step 4. Computing the annealing temperature T. In (8), gen is the iterations, T 0 is the initial temperature, and is the cooling ratio of simulated annealing to control the rate of cooling:
[formula] = gen × 0 .(8) [/formula]
Step 5. The position update of new individual. We divide the neighborhood space for phase and then generate a random update vector , which specifies the location of the phase for updating. One phase is selected randomly to get a new chromosome in the neighborhood space. If the fitness of new chromosome is better than the old one, we will replace the old with the new one. We update the designated quantum bits by through quantum gate according to the above.
Step 6. According to the Metropolis criterion, we update the chromosomes. The probability of new chromosomes acceptation obeys the Boltzmann probability distribution. In [bib_ref] A variant of quantum genetic algorithm and its possible applications, Tiwari [/bib_ref] , fit( ) is the fitness of parent chromosome and fit( ) is the fitness of child chromosome. If fit( ) is greater than fit( ), the new chromosome will be accepted with probability 1. Otherwise, the new chromosome will be accepted with probability :
[formula] ( = ) = { 1 fit ( ) > fit ( ) fit ( ) ≤ fit ( ) , = 1 1 + exp [(fit ( ) − fit ( )) / ] .(9) [/formula]
Step 7. Implement quantum variation operation using the following:
[formula] = 2 − .(10) [/formula]
Step 8. Update the current individuals and execute Step 3 to get the global optimal individual.
Step 9. Determine if it has met the end conditions true is the end to return the optimal parameters and false goes to Step 4.
Step 10. Use the optimal parameters to train an SVM classifier.
After we obtain the classifier using optimal parameters, we will use it to classify the testing samples. Then, we compare the classification labels with the known labels, so that we can get the classification accuracy for evaluating the performance of classifier. It is shown as [bib_ref] Quantum annealing algorithms: atate of the art, Du [/bib_ref]. TN is the total number of
# Results and discussion
We select 114 groups of pancreas images; among them 81 groups are normal and 33 groups are abnormal. The resolution of each image is 128 × 128 and the thickness is 2.77 mm. Among these 114 groups, we select 40 groups of normal images and 16 groups of abnormal images as the testing samples and others as the training samples. shows two images; one of them is normal data and the other one is abnormal data. The pancreatic morphology of abnormal data is thick and big.
We can see the mean of the samples of each 81 normal pancreas and 33 abnormal pancreas in .
In this paper, we use the SVM method to classify the pancreas data. We can see several results in [fig_ref] Table 1: Accuracy of centered data and not centered data using SVM [/fig_ref]. In [fig_ref] Table 1: Accuracy of centered data and not centered data using SVM [/fig_ref] , is the parameter gamma of the kernel function RBF and C is the penalty factor. From [fig_ref] Table 1: Accuracy of centered data and not centered data using SVM [/fig_ref] , we can see that the results on centered data are significantly better than the results on the data without centering, so the variation capture with respect to the data center is more powerful in classification than variation capture with respect to the original data.
The experimental result of 5 groups of QSA-SVM is shown in [fig_ref] Table 2: Experiments result of QSA-SVM [/fig_ref]. We can see that the mean operation duration is approximate 129.35 s, the mean accuracy of classification is 94.64%, the parameter gamma of the kernel function RBF , and the mean values of penalty factor C are 543.12728 and 108.20392.
Compared with other classifiers, the accuracy of QSA-SVM is better which is shown in . And in , the comparison of running time is shown.
Classifier BPNN is BP neural network, the accuracy is 25% and the running time is 6.76 s; classifier Fisher is fisher linear classifier, the accuracy and running time are 35% and 0.98 s; classifier SVM is the common SVM, the accuracy is 71.4286% and the running time is only 0.13 s; classifier ACO-SVM [bib_ref] Share price prediction using wavelet transform and ant colony algorithm for parameters..., Fang [/bib_ref] [bib_ref] Parameters optimization in SVM based on ant colony optimization algorithm, Liu [/bib_ref] is the optimized classifier SVM using ant colony algorithm which has the same accuracy with common SVM, but the running time is 544.51 s, it is much slower than the proposed method. As shows, QSA-SVM has better efficiency in classification than others, and in , it is slower than several methods, but it is faster than ACO-SVM which is also able to optimize the parameters of SVM. At present, radiologists usually diagnose pancreatic diseases with their own experience and the morphology information of image. But missed diagnosis sometimes inevitably happened due to individual differences of patients or limitation of doctor's knowledge of image information. Hence, the proposed method can be used in CAD technology and give early diagnosis of pancreatic diseases in the acceptable time of doctor, so that the classifier can help doctors to diagnose the disease of patient and improve diagnosis rate of disease.
# Conclusions
In this paper, tensors have been used to represent the image and MPCA extended linear PCA to multilinear subspace learning for the tensor object analysis, and QSA-SVM method has been proposed to classify images. As an application for Computational and Mathematical Methods in Medicine 7 classifying pancreatic images, the method combining MPCA and QSA-SVM achieved the better classification accuracy, because MPCA method can preserve the relationship of features in the original tensor and the structure of the original image as much as possible; in the acceptable time, QSA which was used for optimizing SVM classified model is able to find the optimal model parameters. Therefore, the proposed method can improve the classification accuracy of pancreatic images and then assist doctors to diagnose diseases.
[fig] Figure 1: The main flow of the proposed method. [/fig]
[fig] Figure 3: The flow of MPCA algorithm. [/fig]
[fig] For = 1: * set the matrix ( ) to consist of the eigenvectors of the matrix ( ) corresponding to the largest eigenvalues. Calculate { , = 1, . . . , } and Ψ . If Ψ − Ψ −1 < , break and go to next step Pseudocode 1: Pseudocode of the local optimization. [/fig]
[fig] Figure 4: The relationship of the three states. [/fig]
[fig] Figure 5: The flow of QSA-SVM. [/fig]
[fig] Figure 6, Figure 7: Illustration of samples. (a) Normal pancreas. (b) Abnormal pancreas. Illustration of the mean figure of the samples. (a) The mean of the normal samples. (b) The mean of the abnormal samples. [/fig]
[fig] Figure 8, Figure 9: Classification Classification time of pancreas images using 5 classifiers. [/fig]
[table] 2: Computational and Mathematical Methods in Medicine [/table]
[table] Table 1: Accuracy of centered data and not centered data using SVM. [/table]
|
Toxoplasma gondii seroprevalence in the Portuguese population: comparison of three cross-sectional studies spanning three decades
# Introduction
Toxoplasma gondii is an obligate intracellular protozoan parasite belonging to the phylum apicomplexa that infects most mammals worldwide. It undergoes a sexual cycle in the intestinal epithelium of the definitive host, members of the cat family, where it turns into oocysts, which are subsequently shed in the environment. It also undergoes an asexual cycle in intermediate hosts, such as birds, rodents and other mammals including human beings. [bib_ref] Manson's tropical diseases, Holliman [/bib_ref] Oocysts are extremely stable in the environment, and are transmitted to other hosts through inadvertent ingestion. Humans acquire T. gondii through ingestion of tissue cysts in the undercooked meat of intermediate hosts, mainly pork and lamb, or by the ingestion of water or food contaminated by faeces containing oocysts from the definitive host, 2 and rarely through transplantation of an infected organ. [bib_ref] Toxoplasma gondii: transmission, diagnosis and prevention, Hill [/bib_ref] While toxoplasmosis is often benign in immunocompetent individuals (revealing no symptoms or may experience swollen lymph nodes), it induces major complications in immunocompromised individuals and during pregnancy, constituting a life-threatening disease. Congenital toxoplasmosis (transmission to the fetus when a pregnant woman acquires T. gondii infection for the first time during pregnancy) can result in abortion or lead to severe malformation of the fetus or to visual or neurological injuries in the newborn, such as hydrocephalus, cerebral calcification and/or chorioretinitis. Such patients may require prolonged, sometimes
Strengths and limitations of this study ▪ The results of the present study fill an important gap in the Toxoplasma gondii seroprevalence in the Portuguese population as no data were available for the past 30 years. ▪ This study indicates that more than 80% of the Portuguese childbearing women are susceptible to primary infection, yielding a risk of congenital toxoplasmosis and respective sequelae. ▪ Different serological methods were used in the three cross-sectional studies, with possible sensitivity and/or specificity dissimilarities, which could not be overcome due to the chronological distance between them. ▪ The study of risk factors could not be performed due to the lack of more complete personal, socioeconomic and demographic information.
lifelong, therapy. Also, recent studies suggest that subtle behavioural or personality changes may occur in T. gondii-infected humans, [bib_ref] Influence of latent Toxoplasma infection on human personality, physiology and morphology: pros..., Flegr [/bib_ref] and toxoplasmosis has recently been associated with neurological disorders, particularly schizophrenia, 5 and bipolar disorder. [bib_ref] Serum antibodies to Toxoplasma gondii and Herpesvidae family viruses in individuals with..., Tedla [/bib_ref] However, evidence for causal relationships remains limited. [bib_ref] Toxoplasma gondii infection, from predation to schizophrenia: can animal behaviour help us..., Webster [/bib_ref] T. gondii has a wide spectrum of prevalence across the globe and infects up to one-third of the world's population. [bib_ref] Toxoplasma gondii: from animals to humans, Tenter [/bib_ref] Cultural habits in regard to cooking food are most likely the major cause of differences in the frequency of infection with T. gondii in many areas of the world, ranging from <10% to almost 100%. [bib_ref] Toxoplasma gondii: from animals to humans, Tenter [/bib_ref] [bib_ref] Toxoplasma gondii infection in horses. A review, Tassi [/bib_ref] [bib_ref] The prevalence of Toxoplasma antibodies in Guatemala and Costa Rica, Gibson [/bib_ref] [bib_ref] Epidemiology of toxoplasma infections, Feldman [/bib_ref] Consequently, there are large differences in the incidence of congenital infection that may vary from 1:1000 live births in France, to 1:10 000 in countries with a lower seroprevalence, and in the USA, it has been estimated that ∼3500 infants are born each year with congenital toxoplasmosis.According to a recent review, [bib_ref] Toxoplasmosis snapshots: global status of Toxoplasma gondii seroprevalence and implications for pregnancy..., Pappas [/bib_ref] the global status of T. gondii seroprevalence in pregnant women or in reproductive ages ranges from above 60% in countries such as Brazil, Gabon, Indonesia, Germany and Iran, to <10% in the UK and Korea. The surveillance schemes of toxoplasmosis are very heterogeneous in European countries, which hinders the burden of congenital toxoplasmosis estimates comparison.The first National Serological Survey was conducted in continental Portugal between 1979 and 1980 and showed T. gondii overall seroprevalence of 47%. [bib_ref] Inquérito Serológico Nacional, Portugal continental, 1979/80 Prevalência dos anticorpos antitoxoplasmose, Ângelo [/bib_ref] Since then, no National Serological Surveys enrolling T. gondii antibodies were performed, but only small-scale studies. [bib_ref] Legal dispositions and preventive strategies in congenital toxoplasmosis in Portugal, Ângelo [/bib_ref] [bib_ref] Seroepidemiology of Toxoplasma gondii infection in women from the North of Portugal..., Lopes [/bib_ref] Thus, Portuguese health national authorities consider that there is a lack of knowledge of the current epidemiological situation of toxoplasmosis in Portugal, and the 2011 guidelines of the General Directorate of Health for toxoplasmosis screening establish the surveillance of low-risk pregnancy based on three monthly retesting of susceptible women.Of note, congenital toxoplasmosis is a mandatory notifiable disease in Portugal,and its diagnosis should be performed in our laboratory at the National Reference Laboratory of Parasitic and Fungal Infections of the Portuguese National Institute of Health. In this regard, considering our data, there are about three cases of congenital toxoplasmosis per year in Portugal (data not published).
Considering that the unique overall population study performed in Portugal dates back to 35 years ago, the aim of this study was to describe the seroprevalence trends in the Portuguese general population (with special focus on women of childbearing age) over the past three decades , by age group, region and gender.
# Methods
## Study design and sampling
In order to fulfil the defined objectives, three crosssectional seroprevalence studies and 2013) were used. All of them were based on opportunistic sampling. Nevertheless, few biases are associated with this strategy in our study, as sera belonged to individuals seeking diverse blood analysis rather than specific T. gondii evaluation, which also reflects the vast majority of the general population. Furthermore, the same methodology was applied to all three surveys, ensuring their internal validity by increasing their comparability. The starting point was the T. gondii serological data released on behalf of the First Portuguese National Serological Survey in 1979/1980. This sample enrolled 1675 individuals of both genders, which were homogeneously distributed by the following stage groups: 8 months-5 years, 6-15, 16-30, 31-45 and ≥46 years. Each age group included individuals from each of the 18 districts of Portugal (Portuguese islands were not included) that were representative of the population of each district. [bib_ref] Inquérito Serológico Nacional, Portugal continental, 1979/80 Prevalência dos anticorpos antitoxoplasmose, Ângelo [/bib_ref] Then, in this study, we processed two distinct samples. The first one was composed of sera belonging to the Second Portuguese National Serological Survey 2001-2002 that aimed to estimate the prevalence of antibodies to vaccine-preventable diseases, and for this reason the determination of antibodies for toxoplasmosis was not performed at that time. This sample enrolled 3525 individuals of both genders, and covered all age groups. Similar to the first serological survey, each age group included a number of individuals representative of the population of each of the 18 districts. For the participation of individuals, a document was prepared with the objectives and benefits of the study and informed consent was obtained either from the participants themselves or from their legal representatives. [bib_ref] Avaliação do Programa Nacional de Vacinação: 2°Inquérito Serológico Nacional Portugal Continental, Freitas [/bib_ref] From these sera, we used a subsample of 1657 sera, which was established to detect the difference in antibodies prevalence between the 1979/1980 and 2001/2002 Portuguese National Serological Surveys of at least 5%, with a power of 80% and a significance level of 5%. This subsample comprehended individuals of both genders, homogeneously distributed in five age groups: 8 months-5 years, 6-15, 16-30, 31-45 and ≥46 years. Each age group included individuals from each of the 18 districts (as in previous surveys, proportion within districts was ensured). It served as baseline for the establishment of the sample size of the 2013 T. gondii serological survey, which was composed of 1440 individuals of both genders, homogeneously distributed in the same five age groups as the previous survey. In each age group, for both surveys, individuals were distributed by 18 districts of Portugal (Portuguese islands were not included) proportionally to the resident population. For the 2013 survey, individuals were selected (to fulfil the above cited requisites of age, gender and geography) from the attendees of the private clinical laboratory of Dr Joaquim Chaves during the period of January to December 2013. This is the largest Portuguese laboratory comprehending more than 40 sampling collection units dispersed throughout the whole country.
No informed consent was obtained from each participant as, besides the information regarding gender and age, no further information was available to the laboratory. This procedure is in agreement with the Portuguese law No. Serological analyses Sera regarding the 2001/2002 seroprevalence study were analysed for the presence of T. gondii-specific antibodies in 2012, whereas the analysis for the 2013 survey was performed in 2013, both at the National Reference Laboratory of Parasitic and Fungal Infections of the Portuguese National Institute of Health.
T. gondii IgG-specific antibodies were detected by using the automated methodology enzyme linked fluorescent assay-sensibility 99.65% (interval confidence (CI) 94.55% to 97.39%), specificity 99.92% (CI 99.58% to 100%) and cut-off 4≤titre<8 IU/mL, with the VIDAS TOXO IgGII commercial reagents (bioMérieux SA, Marcy-l'Étoile, France) according to the manufacturer's instructions. For the resolution of equivocal samples (ie, 4≤titre<8 IU/mL), we retested them using a manual methodology of the direct agglutination test-sensibility 96.22% (CI 94.55% to 97.39%), specificity 98.80% (CI 96.46% to 99.60%) and cut-off 4 IU/mL, by using the TOXO-Screen DA commercial reagents (bioMérieux SA, Marcy-l'Étoile, France), according to the manufacturer's instructions. Most of the epidemiological surveys are based on the IgG titre because IgG antibody positive titres can be detected 2-3 weeks after infection, reaching a maximum titre within 2 months. It then declines to a baseline level that persists throughout the remainder of one's life. These methodologies were not available at the time the first National Epidemiological Survey was performed (1979/1980), which used an in house indirect immunofluorescence method. [bib_ref] Inquérito Serológico Nacional, Portugal continental, 1979/80 Prevalência dos anticorpos antitoxoplasmose, Ângelo [/bib_ref] Statistical analysis Statistical analysis consisted of the estimation of the seroprevalences among the categories of the variables sex, age groups and region. Differences between the estimated seroprevalences were analysed using the χ 2 test, considering the significance level of 5%. The stats package of R software (V.3.0.3) was used (R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2013. ISBN 3-900051-07-0, URL. http://www.R-project.org).
# Results
In the 2013 serological survey (n=1440), we observed an overall prevalence of T. gondii antibodies of 22% (95% CI 20% to 24%), whereas in the 2001/2002 survey (n=1657) the overall prevalence of T. gondii antibodies was 36% (95% CI 34% to 39%; [fig_ref] Figure 1: A [/fig_ref]. The seroprevalence in the Portuguese population decreased by 11% (95% CI 7% to 14%) between 1979/1980 [fig_ref] Figure 1: A [/fig_ref]. Since there were no statistically significant differences between the 18 districts (data not shown), we grouped them into four regions: north, centre, Lisbon area and south. In the 2013 survey, the seroprevalences ranged from 13% (CI 95% 10% to -17%) in the north of Portugal to 33% (CI 95% 29% to 37%) in the south region [fig_ref] Table 1: Trends of Toxoplasma gondii seroprevalence in Portugal by region calculated with an... [/fig_ref].
This finding is in opposition to the ones observed for the serological surveys of 1979-1980 and 2001/02, where the seroprevalence was higher in the north, centre and Lisbon area. In particular, the north region presented the highest levels in the first two surveys but the lowest in the last survey, in opposition to the south. In general, a decreasing trend was observed for the prevalence of T. gondii antibodies over the studied years in the four regions, with the exception of the south region that showed an increase of 8% (95% CI 2% to 19%) between 2001/2002 and 2013 (tables 1 and 2). In particular, the decrease observed for the north region between these two surveys was statistically significant.
Regarding the association between seroprevalence and age, we observed that the prevalence of specific T. gondii IgG increased significantly with this variable for each age group (eg, for the 2013 survey, χ 2 test for trend in proportions gives a p value <0.001) and it generally decreased over time [fig_ref] Figure 2: Comparison of Toxoplasma gondii seroprevalence trends, according to the data from three... [/fig_ref] Concerning the seroprevalence distribution by gender, we observed no significant differences, either for the 2001/2002 or the 2013 survey. Both genders showed a significant decrease in prevalence between these two surveys [fig_ref] Table 3 Toxoplasma: 95% CI 14% to 22% [/fig_ref].
# Discussion
In this study, we aimed to perform a cross-sectional T. gondii seroprevalence study in the Portuguese population. We acknowledge the following limitations in this serological study, namely: (1) the lack of more complete personal and socioeconomic and demographic information, hampering an evaluation involving risk factors; (2) the probable lower sensitivity and/or specificity of the serological tests used in the first national survey when compared with the ones used in the latest surveys; (3) the use of different serological methods (1979/1980 vs the two most recent surveys); and (4) the use of both fresh (1979/1980 and 2013 surveys) and frozen (2001/ 2002) sera. We believe that some of these limitations could impact some false-negative and/or false-positive results and quantitative data (ie, the determination of serological titres), but less significantly impacted the determination of the immune status of the enrolled people, which was beyond the scope of this study. Also, sensitivity and specificity of the conventional methods most likely changed little over the years, as demonstrated, for instance, by the maintenance of the gold standard (Sabin-Feldman dye test) over the past 60 years. Finally, interassay variability is very unlikely to explain the large changes in seroprevalence observed from 1979/1980 to the 2013 survey.
The 2013 serological survey established an overall prevalence of T. gondii-specific IgG of 22%. This value indicates a prevalence that is similar to what has been described for other Mediterranean countries, such as Spain, Italy and Greece, which could be associated with the similarity between these countries in terms of: (1) climate: seroprevalence may depend on the appropriate conditions for sporulation and oocysts survival in the environment, as oocysts maturation and transmission to a new host is faster at mild and wet climates, [bib_ref] Toxoplasma gondii infection in horses. A review, Tassi [/bib_ref] which are not characteristic of these countries;
(2) cat contact: cats are one of the major sources of infection, as they shed a large amount of oocysts via faeces, indoors and outdoors, even after ingestion of a single bradyzoite or one tissue cyst. More, sporulated oocysts can survive for years in water, gardens, beach sands and farms, which constitute environmental features that are typical of these countries, enabling similar means of transmission due to the contact with the abundant stray and domestic cats in these environments, and (3) diet: The ingestion of raw, undercooked or cured meats is the primary risk factor in Europe for acquiring toxoplasmosis, 1 and these countries present the same nutritional behaviour and eating habits, namely the traditional consumption of cured meats and raw vegetables. Although the frequent cat contact and typical diet would support higher seroprevalences than those observed, other factors, such as the climate and population education (discussed below), most likely balance this trend.
We observed a huge decrease in the T. gondii seroprevalence in Portugal, from 1979/1980 to the 2013 survey, as the first survey revealed a 47% rate, [bib_ref] Inquérito Serológico Nacional, Portugal continental, 1979/80 Prevalência dos anticorpos antitoxoplasmose, Ângelo [/bib_ref] which was low compared with the values reported in France, but sharply high when compared with north European countries. Curiously, the 2013 prevalence (22%) remains low compared with France but is now sharply lower compared with some countries from the north of Europe such as Germany, Poland, Belgium, the Netherlands and Switzerland. [bib_ref] Toxoplasmosis snapshots: global status of Toxoplasma gondii seroprevalence and implications for pregnancy..., Pappas [/bib_ref] We speculate that a possible explanation for those striking differences may rely on cultural habits in regard to eating and cooking practices. The sharp decreasing trend of T. gondii seroprevalence observed in Portugal during the past 34 years has been also reported in many studies performed in several developed countries. [bib_ref] P Prevalence and estimated incidence of Toxoplasma among pregnant women in Poland:..., Nowakowska [/bib_ref] Since the consumption of raw or undercooked meat is considered to be the major source of T. gondii infection, we believe that this decreasing trend may be associated with: (1) the practice of freezing meat; (2) the changes in nutritional habits that took place in Portugal and Europe during the past years related to the wide access to fast foods and pre-prepared meals, including frozen meat meals and the decreasing number of home-prepared meals; [bib_ref] P Prevalence and estimated incidence of Toxoplasma among pregnant women in Poland:..., Nowakowska [/bib_ref] (3) the introduction of intensive farming techniques involving the separation of cats from livestock, coupled with the reduction of breeding cattle in backyards; 1-29 (4) the release of legislation for toxoplasmosis, concerning sanitary inspection in the slaughterhouses; and (5) the improvements in health education and information by health professionals, as the lack of awareness of disease sources of transmission is a crucial factor in the risk of infection.
Regarding the distribution of T. gondii seroprevalence by geographic region, we observed an intriguing phenomenon as, contradicting the continuous decrease detected in the north, centre and Lisbon area, there was an 8% increase (table 1) of the seroprevalence in the south, from 2001-2002 to the 2013 survey. Also, in the 2013 survey, this region was the one with the highest T. gondii seroprevalence (33%), which was significantly higher (about 2.5-fold) than the region showing the lowest values (north-13%). The variation in the geoseroprevalence of this parasite may be due to the local rainy conditions and altitude. [bib_ref] Toxoplasmosis-a global threat. Correlation of latent toxoplasmosis with specific disease burden in..., Flegr [/bib_ref] Nevertheless, this would explain the lower values observed for the south region in the first two surveys when compared with the other regions, but contradicts the scenario observed in the 2013 survey, as the south of Portugal presents a prominent lower altitude and is clearly drier and warmer than the remaining regions.
In most human populations, the T. gondii seroprevalence increases with age, indicating that infection is acquired throughout life. [bib_ref] Manson's tropical diseases, Holliman [/bib_ref] Our study was no exception and showed a higher seroprevalence in older age groups (for all geographical regions), most likely due to their longer exposure to the risk factors.
Concerning childbearing women, we observed a significant decreasing trend in the T. gondii seroprevalence over time [fig_ref] Table 4: Toxoplasma gondii seroprevalence in childbearing women by age group calculated with an... [/fig_ref]. This observation is corroborated by other studies, which show that the prevalence of T. gondii infection in women of childbearing ages has decreased over the past 30 years, [bib_ref] Djurkovic-Djakovic O. Kinetics of Toxoplasma infection in the Balkans, Bobic [/bib_ref] and consequently, more women are now susceptible to the infection. Besides the factors stated above that most likely justify the seroprevalence decrease in the general population, the major factor associated with the decline observed for this target group most likely relies on the promotion of educational programmes yielding an improvement of the primary prevention. The currently observed seroprevalence of 18% indicates that about 80% of Portuguese women are not immune against the T. gondii infection. Thus, the majority of potential pregnant women are susceptible to primary infection and the risk of congenital toxoplasmosis is high, a scenario that mirrors the one observed for several cities in Spain. [bib_ref] Prevalence and incidence in Albacete, Spain, of Toxoplasma gondii infection in women..., Bartolomé Alvarez [/bib_ref] The low prevalence observed among women of childbearing age should not be neglected. In fact, a recent study from the WHO, [bib_ref] The global burden of congenital toxoplasmosis: a systematic review, Torgerson [/bib_ref] which estimated the global burden of congenital toxoplasmosis by using a previously described prediction model, [bib_ref] Models for prediction of the frequency of toxoplasmosis in pregnancy in situations..., Larsen [/bib_ref] alerts that the global burden of the disease is considerably higher than that suggested by the congenital toxoplasmosis data. The incidence of infection is dependent on the general seroprevalence, determining the population susceptibility and the frequency of risk factors for toxoplasmosis acquisition. Therefore, seroprevalence should be considered the indicator to establish screening policies. Regarding this issue, several countries have no surveillance of the infection, whereas others focus solely on severe cases and few have surveillance targeted at congenital toxoplasmosis.In Portugal, there is a surveillance system, which includes both the screening of pregnant women with follow-up during pregnancy of those who are not immune, in order to detect seroconversion, and the mandatory laboratory notification of congenital toxoplasmosis cases detected during this process.This study reinforces that prenatal screening for toxoplasmosis is necessary due to the high rate of seronegative women exposed to infection and the probability of a high number of primary infections in the childbearing period. Targeted information should be crucially provided to childbearing and pregnant women by the health professionals regarding consumption of uncooked or cured meat, raw vegetables, contaminated drinking water and contact with cats. With one exception (Toxovax for sheep), there is no approved vaccine to prevent human or animal toxoplasmosis; therefore, primary prevention is the major tool to prevent the infection in the general population, mainly in high-risk individuals, such as immunocompromised, pregnant and childbearing women. Besides these primary prevention measures, there are several actions that will determine the prevention and control of T. gondii infection, including the governmental inspection of slaughterhouses and food production industries, the improvement of hygienic standards of abattoirs, the promotion of public educational schemes and the establishment of serological screening programmes.
Contributors MJG, BN and JPG developed the study design and were responsible for data management. BN and SS performed the statistical analysis. MJG, IF, AV and SM performed the experimental assays. CC provided the sera samples. MJG and JPG wrote the manuscript.
Funding The Portuguese National Institute of Health funded this study.
Competing interests None declared.
Ethics approval This study was approved by the ethical committee of the Portuguese National Institute of Health.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional data are available.
Open Access This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http:// creativecommons.org/licenses/by-nc/4.0/
[fig] Figure 1: A) shows the evolution of Toxoplasma gondii seroprevalence in Portugal over the past three decades; (B) shows the sample sizes for the three National Serological Surveys, the precise estimated seroprevalences and respective 95% confidence interval (CI). [/fig]
[fig] Figure 2: Comparison of Toxoplasma gondii seroprevalence trends, according to the data from three National Serological Surveys, by age groups. [/fig]
[table] Table 1: Trends of Toxoplasma gondii seroprevalence in Portugal by region calculated with an interval confidence of 95% [/table]
[table] Table 3 Toxoplasma: 95% CI 14% to 22%) in 2013. We also subdivided this population into five groups: 15-20, 21-26, 27-32, 33-38 and 39-45 years and observed that, like in the general population, there was a significant increase of T. gondii seroprevalence with age (except for a unique age group within the 1979-1980 survey), and a significant decreasing trend over time(table 4). [/table]
[table] Table 4: Toxoplasma gondii seroprevalence in childbearing women by age group calculated with an interval confidence of 95% Gargaté MJ, et al. BMJ Open 2016;6:e011648. doi:10.1136/bmjopen-2016-011648 [/table]
|
Quantitative Phosphoproteomics Reveals the Signaling Dynamics of Cell-Cycle Kinases in the Fission Yeast Schizosaccharomyces pombe
Multiple protein kinases regulate cell-cycle progression, of which the cyclin-dependent kinases (CDKs) are thought to act as upstream master regulators. We have used quantitative phosphoproteomics to analyze the fission yeast cell cycle at sufficiently high temporal resolution to distinguish fine-grain differences in substrate phosphorylation dynamics on a proteome-wide scale. This dataset provides a useful resource for investigating the regulatory dynamics of cell-cycle kinases and their substrates. For example, our analysis indicates that the substrates of different mitotic kinases (CDK, NIMA-related, Pololike, and Aurora) are phosphorylated in sequential, kinase-specific waves during mitosis. Phosphoproteomics analysis after chemical-genetic manipulation of CDK activity suggests that the timing of these waves is established by the differential dependency of the downstream kinases on upstream CDK. We have also examined the temporal organization of phosphorylation during G1/S, as well as the coordination between the NDR-related kinase Orb6, which controls polarized growth, and other cell-cycle kinases.
In Brief phosphoproteomics at high temporal resolution to determine the fine-grain differences in substrate phosphorylation timing during the fission yeast cell cycle. This global analysis reveals how multiple different cell-cycle kinases contribute to phosphorylation ordering, as well as the hierarchy of mitotic kinases downstream of the master regulator CDK.
## Data and software availability
# Introduction
As for many cellular processes, kinase-mediated protein phosphorylation plays a major role in regulating the initiation of and progression through the major cell-cycle transitions. Across eukaryotes, cyclin-dependent kinases (CDKs) act as master regulators: driving G1 cells into S phase, blocking re-initiation of DNA replication, and driving G2 cells into mitosis by phosphorylating hundreds of proteins [bib_ref] Global analysis of Cdk1 substrate phosphorylation sites provides insights into evolution, Holt [/bib_ref] [bib_ref] Targets of the cyclin-dependent kinase Cdk1, Ubersax [/bib_ref]. In fission yeast, the timing with which different CDK substrates are first phosphorylated is primarily specified by rising CDK activity traversing a series of sequential substrate-specific activity thresholds, which is consistent with experiments showing that the timing and ordering of different cell-cycle stages can be reorganized by manipulating the activity of a single cyclin-CDK fusion, in the absence of all other cyclin-CDK complexes [bib_ref] Driving the cell cycle with a minimal CDK control network, Coudreuse [/bib_ref] [bib_ref] A single cyclin-CDK complex is sufficient for both mitotic and meiotic progression..., Gutié Rrez-Escribano [/bib_ref].
However, CDK activity is directly responsible for only a subset of the phosphorylation changes during the cell cycle. A number of other protein kinases have been implicated in fission yeast, including Dbf4-dependent kinase (DDK) at G1/S [bib_ref] hsk1+, a Schizosaccharomyces pombe gene related to Saccharomyces cerevisiae CDC7, is required..., Masai [/bib_ref] ; Polo-like, Aurora, and NIMA-related kinases at G2/M [bib_ref] A NIMA homologue promotes chromatin condensation in fission yeast, Krien [/bib_ref] [bib_ref] The fission yeast NIMA kinase Fin1p is required for spindle function and..., Krien [/bib_ref] [bib_ref] The conserved Schizosaccharomyces pombe kinase plo1, required to form a bipolar spindle,..., Ohkura [/bib_ref] [bib_ref] The S. pombe aurora-related kinase Ark1 associates with mitotic structures in a..., Petersen [/bib_ref] [bib_ref] The role of Plo1 kinase in mitotic commitment and septation in Schizosaccharomyces..., Tanaka [/bib_ref] ; the septation initiation network (SIN) kinase pathway (related to the Hippo pathway) during cytokinesis [bib_ref] Pombe's thirteen -control of fission yeast cell division by the septation initiation..., Simanis [/bib_ref] ; and the morphogenesis-related (MOR) kinase pathway (related to the Ndr1/2 pathway) for polarized interphase growth [bib_ref] Fission yeast orb6, a ser/thr protein kinase related to mammalian rho kinase..., Verde [/bib_ref]. CDK is generally thought to function upstream of these other cell-cycle kinases, although the directness and the biological significance of such dependencies are not fully clear [bib_ref] Chemical genetic analysis of the regulatory role of Cdc2p in the S...., Dischinger [/bib_ref] [bib_ref] The fission yeast NIMA kinase Fin1p is required for spindle function and..., Krien [/bib_ref] [bib_ref] The S. pombe aurora-related kinase Ark1 associates with mitotic structures in a..., Petersen [/bib_ref] [bib_ref] The role of Plo1 kinase in mitotic commitment and septation in Schizosaccharomyces..., Tanaka [/bib_ref]. It has been proposed that fission yeast Polo and NIMA kinases also act upstream of CDK as part of a positive feedback loop that boosts their activities as cells enter mitosis [bib_ref] Removal of centrosomal PP1 by NIMA kinase unlocks the MPF feedback loop..., Grallert [/bib_ref] [bib_ref] Centrosomal MPF triggers the mitotic and morphogenetic switches of fission yeast, Grallert [/bib_ref] [bib_ref] Schizosaccharomyces pombe NIMArelated kinase, Fin1, regulates spindle formation and an affinity of..., Grallert [/bib_ref] [bib_ref] Physical and functional interactions between polo kinase and the spindle pole component..., Maciver [/bib_ref].
To understand how these and other kinases function together to bring about progression through the different events of the cell cycle requires a global description of substrate phosphorylation changes in vivo. This is possible using mass spectrometry-based phosphoproteomics, which has been used in studies ranging from yeast to human cells. However, previous studies have lacked the temporal resolution to evaluate the dynamics and physiological role of protein phosphorylation as cells proceed through the major cell-cycle transitions [bib_ref] Absolute proteome and phosphoproteome dynamics during the cell cycle of Schizosaccharomyces pombe..., Carpy [/bib_ref] [bib_ref] Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis, Olsen [/bib_ref] [bib_ref] Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based..., Sharma [/bib_ref]. Here, we present a phosphoproteomicsbased analysis of phosphorylation changes during the cell cycle at a significantly higher temporal resolution than has been previously achieved, using the model eukaryote Schizosaccharomyces pombe (fission yeast). This has allowed us to delineate the different classes of temporal dynamics with which phosphosites are modified during the cell cycle. We also report phosphorylation changes after chemical genetic manipulation of CDK activity to assess the dependencies of other kinases downstream of CDK.
As an example of the value of this dataset as a resource, we have analyzed the substrates of the different mitotic kinases.
This reveals that these kinases are activated downstream of CDK and that the time at which each kinase is activated appears to be determined by the directness of its dependency on upstream CDK activity. We have also examined the temporal ordering of G1/S phosphorylation and the coordination between the nuclear Dbf2-related (NDR) kinase Orb6, which regulates polarized growth, and other cell-cycle kinases. Together, these data highlight the complex network of interactions between different cellcycle kinases and illustrate how they function to order protein phosphorylation during the eukaryotic cell cycle.
# Results
## Phosphoproteomics analysis of the cell cycle at high temporal resolution
Stable isotope labeling with amino acids in cell culture (SILAC) followed by mass spectrometry allows the comparison of protein or phosphorylation levels on a global scale with good quantitative accuracy [bib_ref] A genetic engineering solution to the ''arginine conversion problem'' in stable isotope..., Bicho [/bib_ref]. We have used this approach in the fission yeast S. pombe to analyze the phosphoproteome during the cell cycle. Briefly, cells were released from a G2 arrest, and protein extracts, taken at 20 time points after release, were mixed with extracts from a common heavy labeled culture (Figures 1C and S1A-S1C). Cells synchronized in mitosis were used as the heavy reference to maximize the total number of detected phosphorylation events.
We were able to quantify the relative phosphorylation levels for 10,095 phosphosites but limited all subsequent analysis to 7,298 sites (on 1,578 different proteins) with a localization probability >0.9 . Of these phosphosites, 4,534 are on proteins with human orthologs [bib_ref] PomBase: a comprehensive online resource for fission yeast, Wood [/bib_ref] , 74% of which are on orthologs of proteins reported as being phosphorylated in human cell lines [bib_ref] Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis, Olsen [/bib_ref] [bib_ref] Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based..., Sharma [/bib_ref]. We also (A) Schematic of the SILAC-mass spectrometry workflow to quantify the relative difference in protein or phosphorylation levels between experimental samples and a common reference sample (throughout our experiments, this common reference was collected in mitosis). SILAC involves the differential metabolic labeling of proteins with lysine and arginine isotopes and subsequent discrimination of heavy or light isotope labeled peptides by mass spectrometry, in which the relative ratio between heavy and light peptide intensities is used as a measure of relative abundance. LC-MS/MS, liquid chromatography-tandem mass spectrometry; m/z, mass/charge. (B) Plot of theoretical versus observed H:L SILAC ratios. Heavy and light labeled protein extracts were mixed in ratios spanning a 250-fold range (1:50-50:1). The mean H:L ratio (± SD), calculated from 527 proteins quantified in all 19 mixes, is plotted. SILAC quantifications are approximately linear with expected ratios across a 64-fold range (1:8-8:1). (C) Schematic of the cell-cycle experiment: cells were synchronized by G2 arrest and release and analyzed over the two sequential synchronous cell cycles. The second cycle is $50% as synchronous as the first. See Figures S1A-S1C for details of the experimental design and quantification of cell-cycle progression and synchrony. NETO, new end take off. (D) Histogram of quantified phosphorylation ratios for the 12-min time point, where heavy and light samples were collected in the same conditions and therefore represent an approximate null distribution. Blue dashed lines denote a 2-fold deviation from the expected H:L ratio of 1:1 (Log 2 [H:L] = 0). Median log 2 (H:L) ratio = 0.01, SD = 0.4. More than 98% of values deviate <2-fold from the theoretically expected ratio, supporting the generally high quality of the quantifications reported here. Phosphorylation quantifications are from only singly phosphorylated peptides with a localization probability >0.9. (E and F) Principal-component analysis (PCA) of the relative phosphorylation levels (L:H) of phosphosites after imputation to replace missing values (n = 4,460 phosphosites) (E) and the relative protein levels (L:H) after imputation to replace missing values (n = 3,326 proteins) (F). The first and second components are plotted. Each point corresponds to a single time point during the cell-cycle experiment. Phosphorylation quantifications are from only singly phosphorylated peptides with a localization probability >0.9. See Experimental Procedures for details of PCA and data imputation to replace missing values. . See main text for further description.
(B) Heatmap of the relative phosphorylation levels (L:H) after release from G2 arrest, for phosphosites that change >33 during the time course (n = 1,370). Each row corresponds to a single phosphosite and rows are ordered by hierarchical clustering, after imputation to replace missing values. Values outside the display range are set to the closest extreme. Cell-cycle clusters with clear periodicity are annotated. See [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design and details.
(C) The number of phosphosites in each cell-cycle cluster as defined in [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref]. Percentages are of the total 976 phosphosites in all cell-cycle clusters.
(legend continued on next page) quantified protein level changes at 10 time points during the cell cycle across 3,356 proteins (N.B. S. pombe genome has 5,064 annotated open reading frames [bib_ref] PomBase: a comprehensive online resource for fission yeast, Wood [/bib_ref]. SILAC quantifications from three injections of each sample, each using a different peptide activation method, show strong agreement among one another [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref]. This and the tight distribution of ratios for the 12-min time point, where heavy and light samples were collected under the same conditions [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] , supports the high overall quality of the data presented here.
## The phosphoproteome is highly dynamic and synchronous during the cell cycle
The phosphoproteome over the cell cycle is highly dynamic, with 47.1% of phosphosites changing at least 2-fold (n = 3,439/ 7,298). We restricted our subsequent analysis to values quantified from singly phosphorylated peptides (n = 6,418) to mitigate artifacts caused by phosphorylation of adjacent sites on the same peptide, although both sets of quantifications are provided in . We note that it cannot be excluded that some quantifications are not representative of site-level dynamics, such as phosphosites that involve priming of or from adjacent phosphosites. We then applied an imputation algorithm to populate missing values for all phosphosites with ratios calculated for at least 10 of the 20 time points (n = 4,460) .
If the changes that we have quantified are biologically meaningful, then we would expect strong agreement between adjacent time points, taken from similar points in the cycle. Principal-component analysis (PCA) reveals that phosphorylation changes between adjacent time points are indeed more similar to one another, with the greatest differences being between interphase and mitosis [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref]. In fact, a continuous trajectory through the two-component space can be drawn that perfectly recapitulates the sequence of the time points, which is indicative of high data quality [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref]. In contrast, there is no apparent synchrony within the protein-level quantifications [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] , consistent with previous reports that the proteome is largely stable during the fission yeast cell cycle [bib_ref] A genetic engineering solution to the ''arginine conversion problem'' in stable isotope..., Bicho [/bib_ref] [bib_ref] Absolute proteome and phosphoproteome dynamics during the cell cycle of Schizosaccharomyces pombe..., Carpy [/bib_ref].
We then applied a stringent threshold of a 3-fold change during the experiment (i.e., maximum value >33 minimum value) and used this criterion to define approximately one-third of phosphosites as being very cell-cycle regulated (n = 1,370/4,460) [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref]. These phosphosites are enriched for a wide range of cell-cycle-related Gene Ontology (GO) terms , and 64.4% (n = 882/1,370) are on proteins with annotated human orthologs, the majority of which are on orthologs of human proteins reported as being phosphorylated in a cell-cycle-dependent manner (n = 537/882; [bib_ref] Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis, Olsen [/bib_ref] [bib_ref] Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based..., Sharma [/bib_ref]. By contrast, stable phosphosites (<1.5-fold change) (26.3%, n = 1,869/4,460), are enriched for GO categories related to ribosome biology, target of rapamycin (TOR) signaling, intra-cellular pH homeostasis, RNA splicing, and various small molecule metabolic processes .
## Phosphorylation dynamics during the cell cycle
Clear waves of coordinated phosphorylation and dephosphorylation can be observed among these cell-cycle-regulated phosphosites. We used hierarchical clustering to classify different subsets of sites that showed consistent changes in phosphorylation localized to specific cell-cycle transitions (n = 976/1,370) (Figures 2C-2J; . If the changes in phosphorylation that we have quantified are truly cell-cycle dependent, then the measurements of relative phosphorylation at adjacent time points should be highly related. We calculated the autocorrelation (ACF) for all phosphosites to assess the significance of their phosphorylation changes, based on agreement between quantifications from proximal time points . There is a significant positive autocorrelation between adjacent time points (i.e., lag = 1) for the vast majority of the 976 sites we classified . Furthermore, at higher lag intervals there is a modest negative autocorrelation, as would be expected for sites that are phosphorylated and then dephosphorylated at a subsequent stage of the cycle [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref].
The majority of these cell-cycle-regulated phosphosites fall within the mitotic cluster, although the use of a mitotic reference means the exact proportion of mitotic phosphosites may be an overestimate. As expected, these mitotic sites are enriched for a wide range of GO categories related to mitosis and cell division. A number of unanticipated cellular processes are also enriched, including large ribosomal subunit biogenesis, nuclear pore biology, and RNA transport [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref]. A subset of the sites phosphorylated in mitosis continued to be phosphorylated into G1 (M/G1 clusters), when cytokinesis is also initiated in fission yeast . Consistent with this, the M/G1 phosphosites are enriched for proteins localized to the site of cell division and for proteins involved in rearrangement of the actin cytoskeleton and cytokinetic contractile ring assembly. G1 phosphosites are enriched for proteins localized to the septum, as well as kinesin subunits and regulators of protein guanosine triphosphatase (GTPase) activity [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref].
Progression from G1 into S phase was marked by the phosphorylation of G1/S phosphosites that were then dephosphorylated as cells completed DNA replication, and by the S/G2/M phosphosites, which increased in phosphorylation until the end of the cycle [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref]. S/G2/M phosphosites are enriched for proteins in the replication pre-initiation complex and the replication fork, while G1/S phosphosites are enriched for proteins implicated in the intra-S DNA damage checkpoint and DNA secondary structure binding [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref]. The interphase phosphosites, which increased in phosphorylation as mitosis is completed and became dephosphorylated as cells progressed from G2 into mitosis [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref] , are enriched on proteins related to mRNA processing, including the CCR4-NOT complex. G2 phosphosites [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref] are enriched for proteins involved in cell polarity and lipid biology [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref] , which is consistent with the initiation of polarized growth in early G2 [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref].
The high temporal resolution to these experiments has allowed us to delineate different categories of phosphosites, with different patterns of phosphorylation during the cell cycle, in proteins involved in a wide range of biological processes lists phosphosites in each cluster). We next used this dataset to consider how the repertoire of different cell-cycle kinases establishes these different patterns of phosphorylation during the cell cycle.
## Kinase-specific waves of phosphorylation during mitosis
Mitotic phosphosites are heavily enriched for the CDK consensus sequence [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref] , and we have previously reported that hundreds of ''late'' Cdc2 substrates are directly phosphorylated by Cdc2 as its activity surpasses a series of substrate-specific thresholds in mitosis. However, 40% of mitotic or M/G1 phosphosites do not occur at the minimal CDK consensus sequence (S/TP) and are therefore likely brought about by other mitotic kinases such as the NIMA-related, Polo-like, and Aurora kinases [bib_ref] Polo-like kinases: conservation and divergence in their functions and regulation, Archambault [/bib_ref] [bib_ref] The chromosomal passenger complex (CPC): from easy rider to the godfather of..., Carmena [/bib_ref] [bib_ref] Mitotic regulation by NIMA-related kinases, O'regan [/bib_ref] [bib_ref] The Aurora kinase family in cell division and cancer, Vader [/bib_ref]. Fission yeast has a single version of each of these kinases: Fin1, Plo1, and Ark1, respectively [bib_ref] A NIMA homologue promotes chromatin condensation in fission yeast, Krien [/bib_ref] [bib_ref] The conserved Schizosaccharomyces pombe kinase plo1, required to form a bipolar spindle,..., Ohkura [/bib_ref] [bib_ref] The S. pombe aurora-related kinase Ark1 associates with mitotic structures in a..., Petersen [/bib_ref] [bib_ref] The genome sequence of Schizosaccharomyces pombe, Wood [/bib_ref]. We identified 22 Aurora kinase (Ark1), 80 Polo-like kinase (Plo1), and 28 NIMA-related kinase (Fin1) substrate consensus sites in the mitotic or M/G1 clusters. The Aurora kinase consensus sites are also enriched for previously defined Ark1dependent phosphosites (p = 1.13EÀ7) [bib_ref] Mitotic substrates of the kinase aurora with roles in chromatin regulation identified..., Koch [/bib_ref]. We then used these criteria to define putative substrates of the respective kinase [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref] ; see Experimental Procedures for details). While not all of these will be true substrates, their average behavior can be used to assess the global changes in substrate phosphorylation for each kinase. We note that non-S/TP mitotic phosphosites are enriched for the Polo-like kinase consensus, which is consistent with the larger number of putative Plo1 substrates identified [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref].
To investigate how these different kinases contribute to mitotic phosphorylation dynamics, we compared the phosphorylation of their substrates with Cdc2 substrates. [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref] show the average phosphorylation of Cdc2, Fin1, Plo1, and Ark1 substrates during mitotic entry and mitotic exit. This reveals that the phosphorylation of the substrates of different mitotic kinases peak at different times during mitosis as sequential, overlapping waves. Cdc2 substrates were phosphorylated first, followed by Fin1 substrates and then Plo1 substrates. Ark1 substrate phosphorylation peaked last, about 10 min later in mitosis than Cdc2 substrates (Figures 3B and 3C) (10 min $10% of the cell cycle [Figures S1A-S1C]). At mitotic exit, the dephosphorylation of these sites followed the same pattern: Cdc2 substrates are dephosphorylated first, followed almost immediately by Fin1 substrates and then Plo1 substrates. Finally, Ark1 substrates begin to be dephosphorylated about 10 min after Cdc2 substrates [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref].
These data suggest that the differences in the timing of substrate phosphoregulation during mitosis are brought about, at least in part, by differences in the activity profile of the respective kinases. Importantly, these substrates are not phosphorylated when Cdc2 is inhibited during the initial G2 arrest [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref] , and previous measurements of purified Fin1, Plo1, and Ark1 activity show that the activation of these kinases is blocked when temperature-sensitive alleles are used to inhibit Cdc2 and arrest cells in G2 [bib_ref] The fission yeast NIMA kinase Fin1p is required for spindle function and..., Krien [/bib_ref] [bib_ref] The S. pombe aurora-related kinase Ark1 associates with mitotic structures in a..., Petersen [/bib_ref] [bib_ref] The role of Plo1 kinase in mitotic commitment and septation in Schizosaccharomyces..., Tanaka [/bib_ref]. Taken together, this suggests that these kinases act either directly or indirectly downstream of Cdc2.
## Differential dependency of mitotic kinases on cdk (cdc2)
To investigate the dependency of Fin1, Plo1, and Ark1 on Cdc2 activity, we analyzed the phosphoproteome after cells were exposed to short pulses of different Cdc2 activity levels: cells expressing an ATP analog-sensitive cdc2 allele were arrested with low Cdc2 activity and then released into a range of different 1-NmPP1 concentrations [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref]. When cells are exposed to lower 1-NmPP1 concentrations (i.e., higher Cdc2 activity levels), Cdc2 substrate site phosphorylation increases and a sigmoidal function can be fitted to these data [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref]. Plo1 and Fin1 substrate phosphorylation showed a similar dose response to different levels of CDK activity but were phosphorylated at lower 1-NmPP1 concentrations (i.e., higher Cdc2 activity levels) than direct Cdc2 substrates [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref]. The median 1-NmPP1 half maximal inhibitory concentration (IC 50 ) values are 222, 97, and 79 nM for Cdc2, Fin1, and Plo1 substrate sites, respectively [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref]. These data support the notion that Plo1 and Fin1 activity is regulated directly downstream of Cdc2. By contrast, Ark1 substrate phosphorylation was only partially stimulated at very high Cdc2 activity levels and could not be fitted to a sigmoidal function, consistent with a more indirect downstream dependency. The phosphorylation of Plo1 and Fin1 substrate sites also appeared somewhat more abrupt than direct Cdc2 substrates, as shown by their more negative Hill slopes, with median values of À1.2, À1.5, and À1.6 for Cdc2, Fin1, and Plo1, respectively. This may reflect the possibility that Plo1 and Fin1 activation occurs in a non-linear switch-like manner in response to upstream Cdc2 activity, perhaps ensuring a more stepwise increase in substrate phosphorylation during G2/M.
To further examine these dependencies, we analyzed the phosphoproteome after acute chemical inhibition (10 mM 1-NmPP1) of Cdc2 when Fin1, Plo1, and Ark1 substrates have already become phosphorylated in mitosis [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref]. [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref] shows the average phosphorylation of the substrates of each kinase after Cdc2 inhibition. Cdc2 substrates were dephosphorylated instantaneously after 1-NmPP1 addition, whereas Fin1 substrates showed a small delay of 1-3 min before they began to be dephosphorylated. Plo1 substrates were stable up to $5 min, after which the majority began to be dephosphorylated. By fitting a plateau followed by exponential decay function to the data for individual phosphosites, we calculated the delay interval (X0) between Cdc2 inhibition (0 min) and the time that each site begins to be dephosphorylated. The majority of Cdc2, Fin1, and Plo1 sites could be fitted (see Experimental Procedures for details). [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref] shows the cumulative frequency of X0 values and illustrates the differential responses of these kinases: the median delay interval values are 0.1, 2.4, and 5.0 min for Cdc2 substrates, Fin1 substrates, and Plo1 substrates, respectively. In contrast, Ark1 substrate phosphorylation was refractory to Cdc2 inhibition and even continued to increase after Cdc2 inhibition [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref]. No Ark1 substrate sites could be fitted to the plateau followed by exponential decay function, indicating that Ark1 activity is independent of sustained Cdc2 activity, once cells have entered mitosis.
Taken together, these data indicate that the mitotic kinases have differing dependencies on upstream Cdc2, the strength , means (± SEMs) were calculated from phosphosites that could be fitted to a four-parameter logistic function (see Experimental Procedures for details). No Ark1 substrate sites could be fitted to the model. Means (± SEMs) were normalized to the largest mean during the cell cycle (set to 1.0). Curves are a fourparameter logistic function fit to the means. See [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design. of which is correlated with the timing of their substrate phosphorylation and dephosphorylation during mitosis. Fin1 substrate phosphorylation is more responsive to Cdc2 than Plo1, which in turn is more responsive than Ark1 [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref]. We propose that these differential dependencies are responsible at least in part for the temporal ordering of substrate phosphorylation during mitotic progression. We also quantified a number of phosphosites on these kinases and their regulators, which may act as part of the signaling architecture that brings about this hierarchy of dependencies [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref].
Temporal Organization of G1/S Phosphorylation We next turned to the phosphosites that are regulated during G1 and S phase (G1, G1/S, and S/G2/M clusters, defined in [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref]. In G1, CDK (Cdc2) is inactive before rising to a low level in S phase, resulting in the phosphorylation of a subset of ''early'' Cdc2 substrates, which are first phosphorylated at a significantly lower Cdc2 activity threshold than the ''late'' mitotic substrates. S/G2/M phosphorylation is dominated by the Cdc2 consensus site and includes these early Cdc2 substrates [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref] ; . G1 and G1/S phosphorylation occur in two sequential waves that precede Cdc2 substrate phosphorylation [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref]. This suggests that Cdc2 activation is the final step in the progression from G1 into S phase, in contrast to G2/M where Cdc2 activation precedes that of the other mitotic kinases [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref]. We then asked whether G1 and G1/S phosphorylation showed a dose-dependent response to different levels of Cdc2 activity by analyzing their phosphorylation after arrested cells were exposed to short pulses of different Cdc2 inhibitor (1-NmPP1) concentrations. G1/S phosphosites showed no response to acute Cdc2 activation, consistent with their phosphorylation being a Cdc2-independent event that precedes Cdc2 activation [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref]. Similarly, G1 phosphorylation is not stimulated by Cdc2 activation, but it does increase at very high 1-NmPP1 concentrations, when Cdc2 is completely inactivated [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref]. This suggests that the phosphorylation timing of these G1 phosphosites is brought about downstream of Cdc2 inactivation during G1, although the kinases responsible for this phosphorylation are unknown. The Dbf4-dependent kinase (Hsk1 in fission yeast) operates alongside CDK at G1/S (Gó mez-Escoda and Wu, 2017; [bib_ref] How do Cdc7 and cyclin-dependent kinases trigger the initiation of chromosome replication..., Labib [/bib_ref] , and although multiple Hsk1 substrates have been identified, how Hsk1 activity changes during the fission yeast cell cycle has not been carefully examined. We identified three sites on previously described Hsk1 substrate proteins in the G1/S cluster [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref]. This is consistent with the notion that Hsk1 is activated independently of Cdc2 during G1/S and is supportive of in vitro data from budding yeast, which suggests that efficient DNA replication requires DDK activation to occur before CDK activation.
## Temporal dynamics of growth control-mor pathway
Regulation during the Cell Cycle Finally, we turned to the regulation of cellular growth during the cell cycle. Polarized growth in fission yeast is temporally inhibited during mitosis and cytokinesis, and then resumes as new daughter cells are born in early G2 [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] [bib_ref] Growth in cell length in the fission yeast Schizosaccharomyces pombe, Mitchison [/bib_ref]. The NDR-related kinase Orb6 is part of the MOR kinase pathway that promotes polarized growth in G2 by organizing the polarization of actin patches and regulating the spatial activity of Cdc42 [bib_ref] The conserved NDR kinase Orb6 controls polarized cell growth by spatial regulation..., Das [/bib_ref] [bib_ref] Fission yeast orb6, a ser/thr protein kinase related to mammalian rho kinase..., Verde [/bib_ref] Mean (± SEM) values were calculated from phosphosites that could be fitted to a four-parameter logistic function. Means (± SEMs) were normalized to the largest (set to 1.0) mean during the cell cycle. Curves are a four-parameter logistic function fit to the means. See [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design. (D) Schematic representing the proposed temporal organization of phosphorylation during the G1/S transition: Cdc2 substrate phosphorylation occurs last among the G1/S phosphorylation events rising as at the onset of S phase. 2015), and the different mechanisms linking Orb6 kinase activity to the pattern of polarized growth during the cell cycle are not fully clear. We noticed that the G2 cluster of phosphosites is heavily enriched for the NDR kinase consensus sequence [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref] , indicating that these proteins, which are phosphorylated when Orb6 function is required, may represent direct Orb6 substrates [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref] [bib_ref] Tumor suppressor LATS1 is a negative regulator of oncogene YAP, Hao [/bib_ref]. Consistent with the function of Orb6, these phosphoproteins include the polarity factors Tea3 and Tea5; GTPase regulatory proteins such as Rga3, Rga6, and Rgf3; and a number of proteins involved in lipid transport and membrane deposition .
The SIN, which is activated in late mitosis and promotes cytokinesis [bib_ref] SIN-dependent phosphoinhibition of formin multimerization controls fission yeast cytokinesis, Bohnert [/bib_ref] [bib_ref] The SIN kinase Sid2 regulates cytoplasmic retention of the S. pombe Cdc14-like..., Chen [/bib_ref] [bib_ref] The fission yeast septation initiation network (SIN) kinase, Sid2, is required for..., Feoktistova [/bib_ref] [bib_ref] The kinesin-14 Klp2 is negatively regulated by the SIN for proper spindle..., Mana-Capelli [/bib_ref] , interacts with the MOR pathway because the major SIN kinase Sid2 phosphorylates Nak1 and thereby inactivates Orb6 [bib_ref] Identification of SIN pathway targets reveals mechanisms of crosstalk between NDR kinase..., Gupta [/bib_ref] [bib_ref] The mitosis-to-interphase transition is coordinated by cross talk between the SIN and..., Ray [/bib_ref]. It has been proposed that this crosstalk between the SIN and MOR pathways ensures that Orb6-dependent polarized growth occurs at the appropriate time. To test this, we identified phos- phosites on Sid2 substrates in our dataset [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref] and compared their phosphorylation dynamics to those of the Orb6 substrates. The phosphorylation of Orb6 substrate sites increased after cytokinesis and was maximal throughout G2, when Orb6 activity is required. By contrast, the phosphorylation pattern for Sid2 substrate sites peaked late in mitosis and became dephosphorylated as cells completed cytokinesis [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref]. There is a clear negative correlation between Sid2 and Orb6 activity, but they are not perfectly anticorrelated: early in mitosis SIN pathway activation was still low, while the MOR pathway was already off [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref]. This suggests that while SIN-mediated inactivation of Orb6 is important, there are additional mechanisms to inhibit Orb6 earlier in mitosis. A number of MOR pathway proteins were phosphorylated early in mitosis, including on sites modified by Cdc2, such as Rga4 and Sog2 [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref]. We also quantified a number of phosphosites on SIN pathway proteins that may function to regulate the timing of SIN activation in late mitosis [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref] [bib_ref] Chemical genetic analysis of the regulatory role of Cdc2p in the S...., Dischinger [/bib_ref].
Together, these data suggest that Cdc2 and/or other mitotic kinases crosstalk with the MOR pathway in early mitosis to promote the inhibition of polarized growth, and then as the mitotic kinases are switched off around anaphase, the SIN-mediated inhibition of the MOR pathway takes over [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref]. This represents another example of how crosstalk between different kinase pathways is used to generate coordinated waves of substrate phosphorylation and dephosphorylation during the cell cycle.
# Discussion
Here, we present a global phosphoproteomics analysis during the eukaryote cell cycle. We have analyzed the fission yeast cell cycle at high temporal resolution, allowing us to delineate the different categories of phosphosites, with different temporal patterns of phosphorylation changes [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref]. As an example of how this dataset may be used to investigate different models of cell-cycle control, we have used kinase target consensus sequences to consider the coordination of different cell-cycle kinases [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref].
At G1/S, multiple sequential waves of phosphorylation occur, with CDK substrates being phosphorylated last, acting as the final phosphorylation event driving G1 cells into S phase. Other G1/S substrates are phosphorylated before CDK activation at G1/S, including possible DDK targets [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref]. After cells complete S phase, CDK activity increases abruptly in late G2, phosphorylating a wider range of substrates to initiate mitotic onset. In contrast to G1/S, CDK activation at G2/M precedes the activation of other mitotic kinases. This appears to occur at different times for each kinase, resulting in a series of temporally ordered kinase-specific waves in phosphorylation during mitosis: first CDK, then NIMA-related kinase, then Polo-like kinase, and finally Aurora kinase [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref]. We then used chemical genetic manipulations of CDK to examine the dependencies of these kinases on CDK activity. These experiments indicate that earlier waves are more directly responsive to CDK than later ones. We propose that the timing of mitotic phosphorylation is, at least in part, ordered by the differential dependencies of the different mitotic kinase on upstream CDK activity [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref].
Our analysis also suggests that as early mitotic kinases are activated, the MOR pathway kinase Orb6, which is required for polarized growth during interphase, is switched off. A number of phosphorylation events on MOR pathway proteins suggest that this is caused by CDK and/or other early mitotic kinase(s), thus explaining how polarized growth is shut down at mitotic onset [bib_ref] Growth in cell length in the fission yeast Schizosaccharomyces pombe, Mitchison [/bib_ref]. At the end of mitosis, as the mitotic kinases are inactivated, Orb6 is maintained off by the now-active SIN pathway kinase Sid2, which also promotes cytokinesis [bib_ref] Identification of SIN pathway targets reveals mechanisms of crosstalk between NDR kinase..., Gupta [/bib_ref] [bib_ref] The mitosis-to-interphase transition is coordinated by cross talk between the SIN and..., Ray [/bib_ref]. We propose that these two regulatory pathways account for how polarized growth is shut down in mitosis and is then reinitiated only as new daughter cells are born, but not before [bib_ref] Growth in cell length in the fission yeast Schizosaccharomyces pombe, Mitchison [/bib_ref] [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref]. We note that the MOR and SIN pathways are conserved in higher eukaryotes with the Ndr1/2 and Hippo pathways [bib_ref] Crosstalk between NDR kinase pathways coordinates cell cycle dependent actin rearrangements, Gupta [/bib_ref].
Together, the analyses we present here illustrate how the repertoire of different cell-cycle kinases are coordinated to bring about the timely ordering of substrate phosphorylation during the fission yeast cell cycle, highlighting the role of kinase crosstalk. We anticipate that the extensive dataset of high-temporally resolved phosphorylation dynamics, which we report in Tables S2 and S3, will be of value in investigating different aspects of eukaryotic cell-cycle control more generally, including the possible substrate specificity and temporal regulation of different phosphatases during the cell cycle [bib_ref] A PP1-PP2A phosphatase relay controls mitotic progression, Grallert [/bib_ref].
A significant proportion of the cell-cycle-dependent phosphosites, which we have characterized, are on the orthologs of proteins reported as undergoing changes in phosphorylation during the cell cycle in human cell lines . For example, fission yeast mitotic phosphosites are enriched on the orthologs of human proteins reported as being phosphorylated in mitosis by [bib_ref] Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis, Olsen [/bib_ref] (p = 2.58EÀ5) and [bib_ref] Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based..., Sharma [/bib_ref] (p = 1.35EÀ4). This and the extensive evolutionary conservation of the major eukaryotic cell-cycle regulators in fission yeast [bib_ref] The genome sequence of Schizosaccharomyces pombe, Wood [/bib_ref] suggest that the conclusions and dataset we present here should be directly applicable to investigating cell-cycle phosphoregulation across eukaryotic systems, including Metazoa.
## Experimental procedures
## Data acquisition
The experiments, raw mass spectrometry data, and MaxQuant output files analyzed in this study were first reported by. All of the experiments were performed using SILAC-compatible strains [bib_ref] A genetic engineering solution to the ''arginine conversion problem'' in stable isotope..., Bicho [/bib_ref] in SILAC-adjusted media (Edinburgh minimal medium [EMM] [6 mM ammonium chloride] + 0.25 mg/mL leucine, 0.15 mg/mL uridine, 0.04 mg/mL arginine, and 0.03 mg/mL lysine). Heavy labeled cultures were first grown for >8 generations in SILAC media with heavy arginine (L-arginine:HCl [U13C6, 99%]) and heavy lysine (L-lysine:2 HCl [U13C6, 99%]) isotopes (Cambridge Isotope Laboratories). Protein extracts from light labeled experimental samples are mixed with protein extract from heavy labeled cells collected in mitosis and vice versa. The design of each experiment is detailed in [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref].
Sample preparation, data acquisition, and processing of the raw mass spectrometry data in MaxQuant (version 1.3.0.5) [bib_ref] MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide..., Cox [/bib_ref] are detailed by. Briefly, heavy and light labeled samples were mixed in an $1:1 ratio, digested with trypsin and Lys-C, and enriched for phosphopeptides using titanium dioxide. Flowthrough fractions were retained for analysis of non-phosphorylated peptides and separated into 12 fractions using strong cation exchange (SCX) liquid chromatography. A linear trap quadropole (LTQ)-Orbitrap Velos was used for data acquisition of phosphopeptides, and an LTQ-Orbitrap Velos Pro (Thermo Scientific) was used for data acquisition of non-phosphorylated peptides. Both instruments were coupled to UltiMate 3000 high-performance liquid chromatography (HPLC) systems (Thermo Scientific) for online liquid chromatographic separation. Phosphopeptide mixtures were injected three times, with one activation method per run: collisioninduced dissociation (CID), multistage activation (MSA), and higher-energy collision dissociation (HCD). Non-phosphopeptide mixtures were injected three times, with CID used as the activation method.
## Data structure and processing
Datasets from four experiments were analyzed in this study. All raw data and MaxQuant output files can be accessed via the PRIDE partner repository (http://www.ebi.ac.uk/pride/archive/) with the accession number PRIDE: PXD003598. The following experiment identification numbers can be used to retrieve the associated data from each experiment:
1. Cell-cycle proteome (experiment ID: CCC6442 [part of the CCC6442
CCC6608 upload], see also ; see [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design) 2. Cell-cycle phosphoproteome (experiment ID: CCC6254 [part of the CCC6061 CCC6254 CCC6758 upload], see also Tables S2 and S3; see [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design)
3. 1-NmPP1 dose-response phosphoproteome (experiment ID: CCC7380 [part of the CCC6879 CCC7380 upload], see also ; see [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design).
For all of the analyses presented here, MaxQuant output files were imported into Perseus (version 1.4.0.2), and normalized heavy-to-light (H:L) ratios were used. Other statistical values (e.g., posterior error probability, localization probability) were also imported directly from MaxQuant output files. For phosphopeptides, reverse and contaminant peptides were removed, and phosphosites with >0 valid values quantified were retained. All analyses described below used data exclusively from the quantification of only singly phosphorylated peptides with a localization probability >0.9, although quantifications from both all phosphopeptides and only singly phosphorylated peptides are provided in, and S5.
## Imputation to replace missing values and data smoothing
Imputation was applied to replace missing values (NaN) using the R package DMwR (R Foundation). Missing values were replaced by a weighted average of the corresponding time point values of 10 nearest neighbors (weights given by the calculated Euclidean distance) using default parameters of the 'knnImputation' function (10 nearest neighbors, k-nearest neighbors algorithm for selecting the neighbors). Only phosphosites or protein groups with SILAC ratios calculated for at least half of the time points were retained and processed for imputation. Imputation was applied only to quantifications from the cell-cycle experiment (CCC6254 and CCC6442); 10.6% of the ratios in the CCC6254 (9,471/89,200) dataset and 0.94% of the CCC6442 dataset (529/33,260) constituted imputed values after imputation was applied. Data smoothing was applied after data imputation using a custom R script to substitute every value with the mean value of the five nearest neighbors identified by Euclidian distance. Smoothed and imputed data for the cell-cycle experiment are listed in alongside the number of values imputed for each site. Imputed or smoothed data are presented or analyzed only where stated. All splines were calculated in Prism 6 using default settings.
PCA and Hierarchical Clustering PCA [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] and hierarchical clustering [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref] were performed in Perseus version 1.4.0.2 after imputation to replace missing values (see above for details on imputation). The PCA settings were default. The clustering settings were cluster rows, Euclidian distance, and do not presuppose K-means.
## Acf calculation
The ACF was calculated using the base R function acf. (default settings) for each phosphosite after imputation was applied to replace missing values (see above for details on imputation). The 95% confidence interval in [fig_ref] Figure 2: Cell-Cycle-Dependent Phosphorylation [/fig_ref] is Bartlett's formula on an uncorrelated series and is taken from the R function plot.acf. ACF values with lag = 1 are listed for each site in . We note that autocorrelation analysis is not adjusted for the variation in time intervals between time points in our experiments.
## Definition of (putative) kinase substrate sites
The sites defined as putative substrates of each respective kinase are listed in . Phosphosites in the Mitotic or M/G1 cluster that conform to the Plo1, Fin1, or Ark1 target consensus sequence (defined in [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref] were classified as putative substrates of the respective kinase [bib_ref] Spatial exclusivity combined with positive and negative selection of phosphorylation motifs is..., Alexander [/bib_ref] [bib_ref] Phospho-regulation of kinetochore-microtubule attachments by the Aurora kinase Ipl1p, Cheeseman [/bib_ref] [bib_ref] Quantitative phosphoproteomics identifies substrates and functional modules of Aurora and Polo-like kinase..., Kettenbach [/bib_ref] [bib_ref] Mitotic substrates of the kinase aurora with roles in chromatin regulation identified..., Koch [/bib_ref] [bib_ref] Molecular basis for the substrate specificity of NIMA-related kinase-6 (NEK6). Evidence that..., Lizcano [/bib_ref] [bib_ref] Identification of a consensus motif for Plk (Polo-like kinase) phosphorylation reveals Myt1..., Nakajima [/bib_ref] [bib_ref] Use of an oriented peptide library to determine the optimal substrates of..., Songyang [/bib_ref]. Ark1 substrates were additionally filtered for sites previously described as being phosphorylated in an Ark1-dependent manner [bib_ref] Mitotic substrates of the kinase aurora with roles in chromatin regulation identified..., Koch [/bib_ref] (denoted Ark1* in . For the analyses presented in [fig_ref] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis [/fig_ref] , Cdc2 substrates were defined as sites in the mitotic cluster previously described as late Cdc2 targets. For the analyses presented in [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref] , Cdc2 substrates were defined as sites in the S/G2/M cluster previously described as early Cdc2 targets. Putative Hsk1 substrates were defined as phos-phosites in the G1/S cluster on proteins previously described as being Hsk1 targets [fig_ref] Figure 4: Phosphoregulation during G1/S [/fig_ref] [bib_ref] Hsk1 kinase and Cdc45 regulate replication stress-induced checkpoint responses in fission yeast, Matsumoto [/bib_ref] [bib_ref] Hsk1-and SCF(Pof3)-dependent proteolysis of S. pombe Ams2 ensures histone homeostasis and centromere..., Takayama [/bib_ref]. Putative Sid2 substrates were defined as phosphosites that conform to the Sid2 target consensus sequence on previously described Sid2 substrates [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref] [bib_ref] SIN-dependent phosphoinhibition of formin multimerization controls fission yeast cytokinesis, Bohnert [/bib_ref] [bib_ref] The SIN kinase Sid2 regulates cytoplasmic retention of the S. pombe Cdc14-like..., Chen [/bib_ref] [bib_ref] The fission yeast septation initiation network (SIN) kinase, Sid2, is required for..., Feoktistova [/bib_ref] [bib_ref] Identification of SIN pathway targets reveals mechanisms of crosstalk between NDR kinase..., Gupta [/bib_ref] [bib_ref] The kinesin-14 Klp2 is negatively regulated by the SIN for proper spindle..., Mana-Capelli [/bib_ref]. Putative Orb6 substrates were defined as phosphosites in the G2 cluster that conform to the NDR kinase target consensus sequence [fig_ref] Figure 5: Crosstalk between SIN and MOR Kinases [/fig_ref] [bib_ref] Tumor suppressor LATS1 is a negative regulator of oncogene YAP, Hao [/bib_ref].
## Phosphosite consensus sequences
The amino acid distribution surrounding the phosphosites was analyzed using iceLogo (p = 0.01) [bib_ref] Improved visualization of protein consensus sequences by iceLogo, Colaert [/bib_ref] with sequence windows exported from Perseus version 1.4.0.2. Each cluster of cell-cycle phosphosites was compared to the background dataset of the 4,460 phosphosites listed in . . Annotation categories enrichment for non-S/TP phosphosites was determined as above but using a background dataset of non-S/TP phosphosites (n = 2,954). All significantly enriched GO categories are listed in Tables S7 and S8. Enrichment analysis of putative Plo1 and Fin1 substrate sites on previously identified CDK substrate proteinsand the enrichment of putative Ark1 substrate sites on previously defined Ark1-dependent phosphorylation sites [bib_ref] Mitotic substrates of the kinase aurora with roles in chromatin regulation identified..., Koch [/bib_ref] were performed as described above using a background dataset of all non-S/TP phosphosites (n = 2,954).
## Go and annotation enrichment analysis
## Comparison with human phosphoproteomics datasets
Published data from two previous studies were used for comparison with human cell-cycle-dependent phosphoproteins [bib_ref] Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis, Olsen [/bib_ref] [bib_ref] Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based..., Sharma [/bib_ref]. Phosphosites in human proteins that peak in phosphorylation at G1, G1/S, early S, late S, G2/M, or mitosis identified by [bib_ref] Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis, Olsen [/bib_ref] and phosphosites that increase in mitosis and have R50% phosphosites stoichiometry identified by [bib_ref] Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based..., Sharma [/bib_ref] were used to define human proteins that are phosphorylated in a cell-cycle-dependent manner. The curated S. pombe protein human ortholog list was downloaded from PomBase (https://www.pombase.org) [bib_ref] PomBase: a comprehensive online resource for fission yeast, Wood [/bib_ref] and used to generate an annotation list of S. pombe proteins with human orthologs that are phosphorylated in a cell-cycle-dependent manner. lists all S. pombe phosphosites that are cell-cycle dependent and have a human ortholog with a reported cell-cycle-dependent phosphosite or phosphosites.
Custom Perseus annotation files were built of S. pombe proteins with orthologs that contain a phosphosite or phosphosites that peak in mitosis and has R50% phosphosites stoichiometry identified by [bib_ref] Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis, Olsen [/bib_ref] and S. pombe proteins with orthologs that contain a phosphosite or phosphosites that increase 3-fold in mitosis and has R50% phosphosites stoichiometry identified by [bib_ref] Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based..., Sharma [/bib_ref]. Annotation enrichment analysis was then performed as above (see GO and annotation enrichment analysis) to determine the enrichment of S. pombe mitotic phosphosites on the orthologs of human mitotic phosphoproteins.
## 1-nmpp1 dose-response calculation
The relative phosphorylation (L:H) of Cdc2, Ark1, Plo1, or Fin1 substrate sites were fitted to a four-parameter logistic function (Prism 6). The curve was constrained such that the Hill slope was <0 and Bottom was >0. Outlier detection was used (Q = 10%). Initial values were chosen automatically and all other settings were default. IC 50 and Hill slope were calculated for those phosphosites that could be adequately fitted to the function (i.e., R 2 >0.9, Bottom <0.5, Top <2, and Top/Bottom >2). (Experiment ID: CCC7380, see also; see [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design.)
## Dephosphorylation kinetics calculations
The relative phosphorylation (L:H) of Cdc2, Ark1, Plo1, and Fin1 substrate sites after Cdc2 inhibition in mitosis was fitted to the plateau followed by one-phase decay function (Prism 6). The curve was constrained such that the X0 > 0 and Y0 > 0.5, plateau > 0 and K > 0. Outlier detection was used (Q = 10%). Initial values were chosen automatically, except for X0 (X0[initial] = 0.1). All other settings were default. X0 values were calculated for those phosphosites that could be adequately fitted to the function (i.e., R 2 > 0.7, plateau > Y0). (Experiment ID: CCC5978, see also ; see [fig_ref] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal... [/fig_ref] for experimental design.)
## Data and software availability
The mass spectrometry proteomics data used in this paper was first reported in. All raw data and MaxQuant output files can be accessed via the PRIDE partner repository (http://www.ebi.ac.uk/pride/ archive/) with the accession number PRIDE: PXD003598.
## Supplemental information
Supplemental Information includes five figures and eight tables and can be found with this article online at https://doi.org/10.1016/j.celrep.2018.06.036.
[fig] Figure 1: SILAC-Based Proteomics Analysis of the Fission Yeast Cell Cycle at High Temporal Resolution [/fig]
[fig] Figure 2: Cell-Cycle-Dependent Phosphorylation (A) The number of phosphosites at each stage of analysis applied to the cell-cycle experiment dataset. The initial dataset contains all phosphosites with at least one L:H ratio quantified [/fig]
[fig] Figure 3: The Dynamics of Kinase-Specific Phosphorylation during Mitosis (A) The criteria used to define Cdc2 (CDK) and putative Fin1 (NIMA-related kinase), Plo1 (Polo-like kinase), and Ark1 (Aurora kinase) substrate sites during mitosis (see Experimental Procedures for details). Cdc2 substrates presented in this figure are only for those phosphorylated in mitosis defined as late substrates by Swaffer et al. (2016). (B-D) The normalized mean (± SEM) phosphorylation (L:H) of Cdc2, Fin1, Plo1, and Ark1 substrate sites during mitotic entry (B and C) and mitotic exit (D). Mean (± SEM) values between 0 and 60 min (B and D) or 50 and 130 min (C) were normalized to the smallest (set to 0.0) and the largest (set to 1.0) values. Spline connects points. See Figures 1C and S1A for experimental design and details. (E) The normalized mean (± SEM) phosphorylation (L:H) of Cdc2, Fin1, Plo1, and Ark1 substrate sites at different 1-NmPP1 concentrations. For Cdc2, Fin1, and Plo1 [/fig]
[fig] Figure 4: Phosphoregulation during G1/S (A) The criteria used to define Cdc2 (CDK) and putative Hsk1 (DDK) substrate sites during the G1-to-S transition (see Experimental Procedures for details). Cdc2 substrate sites presented in this figure are only for those in the S/G2/M cluster defined as early substrates by Swaffer et al. (2016). (B) The normalized mean (± SEM) phosphorylation (L:H) of G1 phosphosites, G1/S phosphosites, and Cdc2 and Hsk1 substrate sites during the G1-to-S transition. Mean and SEM values during G1/S (20 and 50 min) were normalized to the smallest (set to 0.0) and the largest (set to 1.0) mean value. Spline connects points. See Figures 1C and S1A for experimental design and details. (C) The mean (± SEM) phosphorylation (L:H) of G1 phosphosites, G1/S phosphosites, and Cdc2 substrate sites at different 1-NmPP1 concentrations. [/fig]
[fig] Figure 5: Crosstalk between SIN and MOR Kinases (A) The criteria used to define putative Sid2 and Orb6 substrate sites (see Experimental Procedures for details). (B and C) The normalized relative phosphorylation (L:H) of individual Orb6 substrate sites (B) and Sid2 substrate sites (C) during the cell cycle. Data imputation and smoothing were applied before values were normalized to the smallest (set to 0.0) and the largest (set to 1.0) values. Spline connects points. See Figures 1C and S1A for experimental design and details. (D) The normalized mean (± SEM) phosphorylation (L:H) of Sid2 and Orb6 substrates during the cell cycle. Mean (± SEM) values were normalized to the minimum (set to 0.0) and maximum (set to 1.0) values. Spline connects points. See Figures 1C and S1A for experimental design and details. (E) Plot of SIN pathway signaling against MOR pathways signaling (i.e., normalized mean phosphorylation values for Sid2 and Orb6 substrate inFigure 5D). Each point corresponds to an individual time point. Time points from early mitosis are the dark gray values and labeled with their time after release (min). (F) Schematic representing the proposed crosstalk between different kinase pathways for the control of Orb6 activity (MOR pathway). Mitotic kinases (e.g., Cdc2) directly or indirectly inactivate Orb6 in early mitosis before Sid2 (SIN pathway)mediated inhibition during late mitosis and cytokinesis. [/fig]
|
Differentiation of human induced pluripotent stem cells to authentic macrophages using a defined, serum-free, open-source medium
Human induced pluripotent stem cells (iPSCs) and macrophages derived from them are increasingly popular tools for research into both infectious and degenerative diseases. However, as the field strives for greater modeling accuracy, it is becoming ever more challenging to justify the use of undefined and proprietary media for the culture of these cells. Here, we describe a defined, serum-free, open-source medium for the differentiation of iPSC-derived macrophages. This medium is equally capable of maintaining these cells compared with commercial alternatives. The macrophages differentiated in this medium display improved terminally differentiated cell characteristics, reduced basal expression of induced antiviral response genes, and improved polarization capacity. We conclude that cells cultured in this medium are an appropriate and malleable model for tissue-resident macrophages, on which future differentiation techniques can be built.
# Introduction
Resident phagocytes are an evolutionarily conserved cell type in metazoans. In mammals, resident macrophages support tissue homeostasis through a wide range of specialized trophic, remodeling, and defense functions, whose importance is illustrated by their failure in malignant, degenerative, and infectious diseases. Numerous methods for the differentiation of macrophages from stem cells have been developed over the years, building upon advances in our understanding of the requirements for hematopoiesis. These can be broadly categorized into three types: monolayer cultures, co-cultures with xeno cells, and via embryoid body (EB) intermediates. Each vary in number of growth factors applied, timing, complexity, yield, and cost. A summary of these methodologies can be found in . To investigate the molecular pathways involved in the pathogenesis of both infectious diseases, such as those caused by human immunodeficiency virus-1 (HIV-1)and Mycobacterium tuberculosis, and degenerative diseases, especially Parkinson diseaseand Alzheimer disease, we developed a pathophysiologically authentic, yet genetically tractable, model of human tissue macrophages derived from pluripotent stem cells. We have shown that this model resembles an MYB-independent early wave of myelopoiesis in vivo that gives rise to resident mac-rophages in a number of tissues. Our first method involved spontaneous differentiation of mesoderm from embryonic stem cells via EBs, followed by myeloid differentiation using macrophage colony stimulating factor (M-CSF) (CSF-1) and interleukin-3 (IL-3) in a serum-supplemented medium. Since this was not effective for all pluripotent stem cell lines, and to remove the undefined serum component, we subsequently developed a serumfree method, using BMP4, VEGF, and SCF to promote mesodermal lineage and prime hemogenic endothelium differentiation during EB formation, and the serum-free X-VIVO 15 (XVIVO) medium (Lonza) during the M-CSF/ IL-3-directed myeloid differentiation stage (van.
With a greater focus than ever on ensuring research integrity in the life sciences through the adoption of open practices in data curation, publication, and the sharing of materials; National Academy of Sciences, National Academy of Engineering (US) and Institute of Medicine (US) Committee on Science, Engineering, and Public Policy, 2009), it is desirable to replace proprietary media used for culturing and differentiating human pluripotent stem cells with ones that are both defined and fully open-source. While great improvements have been made to move away from xenomaterial containing co-culture and serum-based methodologies toward well-defined monolayer and EB protocols, of the methods described in , only one has been described to utilize fully defined and open-sourced media throughout differentiation. The remainder either supplement cultures with serum, or rely on commercial, serum-free alternatives of undisclosed composition. The suppliers of both the commercial induced pluripotent stem cell (iPSC) culture medium, based on previously published medium, TeSR-E8 (STEMCELL Technologies), and X-VIVO 15 (Life Technologies) decline to publish the composition of these optimized materials on commercial grounds. Consequently, the scientific community does not know whether components that may have a material effect on macrophage physiology are present, nor their concentration. Examples could include, but are not limited to, antiinfectives, anti-inflammatories, cell-permeable iron-chelators cryoprotectants, and excessive concentrations of glucose. Accordingly, we have sought to base replacement media on widely available and open-source materials, enabling others to reproduce our experiments conveniently and without undue dependence on particular suppliers.
In this paper, we describe OXM, an open-source alternative macrophage differentiation medium based on Advanced DMEM/F-12 (aDMEM/F-12). We show that OXM generates homogeneous macrophage precursors very comparable to those produced by our earlier methods, both phenotypically and transcriptionally. We note that these cells show signatures consistent with more complete terminal differentiation, including improved morphology and cell-cycle arrest, and have lower basal expression levels of interferon-inducible genes while being more responsive to inflammatory stimulation. This method therefore generates cells that are even closer models of the ''surveilling'' state of homeostatic tissue macrophages.
# Results
XVIVO medium, but not OXM medium, contains undisclosed molecules We first developed a serum-free, defined, open-source medium, named OXM, for differentiation of iPSCs to macrophages. Previously, we described methodologies for differentiationin which Advanced DMEM or RPMI supplemented with 10% fetal calf serum (FCS), or a serum-free alternative, XVIVO, were used to culture the cells. However, the composition of XVIVO is proprietary, we therefore could not know whether it contained additives that may affect the phenotype of cells differentiated in this medium. We therefore decided to use aDMEM/F-12 buffered with HEPES as the base of OXM. To replace FCS, additional insulin and tropolone, a cell-permeable iron chelator shown to be a suitable substitute for transferrin in cell culture (Field, R.P., Lonza Group, 2003. Animal cell culture. US Patent 6593140), were included. M-CSF and IL-3 were supplemented independently. For full composition, see .
To identify possible causes of differences between XVIVO-and OXM-cultured macrophages, we carried out a limited investigation into the chemical composition of each medium using high-resolution negative-ion liquid chromatography-mass spectrometry (LC-MS), specifically screening for metabolites and polar molecules, but not proteins or non-polar molecules. A comparison of the total ion chromatograms (TIC) shows that there are differences in the chemical composition between OXM and XVIVO , as indicated by the intensity of individual peaks. Each peak was analyzed and putatively assigned as described in the supplemental experimental procedures. A list of predicted metabolites is given in the supplemental information . Specifically, we noted a 2-fold difference in C6 sugar concentration between XVIVO and OXM. We subsequently independently confirmed this by enzymatic assay, finding 34.3 mM glucose in XVIVO and 16.7 mM in aDMEM/F-12 . LC-MS also predicted that the levels of amino acids were at higher concentrations in OXM than in XVIVO except for tryptophan (44 nM in OXM, 6.3 mM in XVIVO) . In addition, we noted that three peaks in the TIC of the XVIVO media were absent in that of the OXM media . The best predicted molecules for these peaks are isolariciresinol 9 0 -Oalpha-L-arabinofuranoside (CID: 131751348, , methyl glucosinolate (CID: 9573942, and dithinous acid (CID: 24490, . These molecules were either absent from the OXM media or below the detection limit.
Differentiation of iPSCs to macrophages in novel fully defined, open-source medium, OXM We next sought to compare the maturation states of OXMdifferentiated and cultured macrophages against those differentiated and cultured in XVIVO. EBs were formed and cultured for 4 days according to our previous publication. Minor modifications to methodology were made, including culturing in an inhouse, defined, open-source medium derived from E8we call OXE8. Note that, while the published composition is open-source, commercial media based on E8 have been modified and have proprietary compositions. OXE8 contains FGF-2 and transforming growth factor b and further supplements of VEGF, BMP4, and SCF (for composition see . After EB formation, as in our earlier methods, EBs were transferred into OXM-or XVIVO-based differentiation medium supplemented with IL-3 and M-CSF. For terminal macrophage differentiation media, IL-3 and tropolone were then omitted; . Morphological differences between differentiation cultures were immediately apparent. EBs produced larger cyst-like structures and greater adherent stromal growth in OXM medium. By week 3 of differentiation, monocyte-like macrophage precursors cells (Pre-Macs) had started to be released into the supernatant. This continued for >16 weeks. Interestingly, OXM differentiation cultures produced a large number of PreMacs in the first 3 weeks of production before their yield reduced to 7-8 3 10 4 cells/mL/week. XVIVO differentiation cultures were slower to yield PreMacs, but by week 6 produced approximately 2-fold more cells, an average of 1.4 3 10 5 cells/mL/week. After 16 weeks the average total yield was (4.48 ± 1.26) 3 10 7 cells in OXM and (7.99 ± 2.61) 3 10 7 cells in XVIVO. PreMacs produced in OXM were significantly smaller than in XVIVO (range 15.8-19.5 mm, mean 17.9 mm, versus 13.8 to R20 mm, mean 19.5 mm, respectively). Note that owing to limitations of the NucleoCounter NC-3000, 20 mm is the upper limit of cell size, so XVIVO-produced cells may have been larger.
For terminal differentiation of PreMacs into macrophages, cells were cultured for a further 7 days in macrophage differentiation medium. After 7 days, XVIVOcultured cells remain rounded or acquire a bipolar morphology, with large vesicle-like structures visible. OXM-cultured cells were morphologically heterogeneous with rounded cells, large, flattened cells, and cells with multiple projections
## Weekly yield
Week resuspension. OXM-cultured macrophages were significantly smaller than those in XVIVO: mean 17.51 mm (range 16.28-18.53 mm) versus mean 19.85 mm (range 19.35 to R20 mm). Differences in time taken to resuspend OXM versus XVIVO cells showed that OXM cells were more adherent, both when cultured in macrophage medium or in fresh differentiation medium . In summary, OXM supports the production of more adherent, smaller macrophages than XVIVO, albeit with a lower cumulative yield.
Macrophages cultured in OXM are phenotypically similar to, but distinct from, XVIVO-cultured cells To determine macrophage phenotype, surface marker abundance was determined by flow cytometry. Key markers of macrophage lineage, the lipopolysaccharide (LPS) co-receptor CD14, and the pan-leukocyte marker CD45, were highly expressed in macrophages derived using both media, with no significant differences. Nor was a difference observed with the expression of scavenger receptor CD163. Levels of the Fcg immunoglobulin G (IgG) receptor, CD16, were significantly higher in OXM-cultured cells. Conversely, both the b2-integrin receptor, CD11b, and the costimulatory receptor, CD86, were more highly expressed in XVIVO-cultured cells, suggesting a more M1-like phenotype . Consistent with our previous reports, major histocompatibility complex class II (human leukocyte antigen-DR [HLA-DR]) expression is low in these unstimulated cells in either condition . Expression of the tissue-resident macrophage marker CD68 was observed intracellularly but not on the cell surface, consistent with earlier literature, and was significantly higher in OXM-cultured cells (Figures 3H and S1H).
PreMac cells gave broadly similar results to those of fully differentiated macrophages, although with differences in CD14 and CD16 expression. Overall, the Pre-Macs and macrophages produced using both OXM and XVIVO media display a macrophage phenotype, but show small and consistent differences that may reflect differences in polarization.
Finally, to test the phagocytic capacity of macrophages produced under the two conditions, we measured phagocytic uptake of Alexa 488-conjugated zymosan, a yeastderived particulate glycan. Phagocytic uptake after 30 min was not significantly different between conditions across three genetic backgrounds, indicating that macrophages cultured in OXM are phagocytically competent.
While transcriptionally very similar, OXM-cultured macrophages have a more homeostatic, and XVIVOcultured macrophages a more immunologically active, signature We used RNA sequencing (RNA-seq) to investigate the detailed expression profile of our iPSC-derived tissue macrophages. We first compared the macrophages cultured in XVIVO or OXM to a published dataset of iPSC-derived human microglia (tissue-resident macrophages of the brain) that had been differentiated via induced hematopoietic progenitor cells (iHPCs) derived using modified protocols from previous reports. Principal-component analysis , in which PC1 explained 27% of the variance, and PC2 explained 18%, showed tight clustering of OXM-and XVIVO-cultured cells together. The iPSCderived macrophages showed closest similarity to microglia rather than iHPCs or monocyte-derived macrophages (MDMs), and least similarity to pluripotent cells and definitive hematopoietic lineages, such as dendritic cells . This was expected because iPSC macrophages have a primitive ontogeny, similar to that of microglia, which develop from primitive, yolk sac/embryonic fetal macrophages, unlike adult definitive MDMs that develop from hematopoietic progenitors in the bone marrow. Further differential expression analysis found 15,951 genes shared in both OXM and XVIVO populations, and 2,335 genes unique to either OXM (1,331) or XVIVO (1,004). Analysis of the sources of this variation showed that the media used explained approximately 30% of the variation between samples, while the age of the differentiation culture (as indicated by the week of PreMac harvesting) accounted for approximately 15%, and the remainder was due to residual, undefined sources.
The three genes most significantly upregulated in OXM versus XVIVO were the carbohydrate binding lectin, CHI3L1, the tetraspanin, CD37, and the protease, HTRA4 (p = 0.0028, 0.0049, and 0.006, respectively). In XVIVO, the most upregulated genes were the interferon-inducible protein IFIT3, the endonuclease RNASE1, and the epoxidase SQLE (p = 0.0028, 0.006, and 0.006, respectively). The most enriched gene ontology (GO) term in XVIVO compared with OXM-cultured cells, was ''response to virus''. Multiple interferon-inducible genes, e.g., MX2, RSAD2, and the IFIT family of proteins, were more highly expressed in XVIVO-cultured cells. How-ever, some antiviral receptors, such as TLR8, are more highly expressed in OXM-cultured cells. We also found enrichment of metabolism-related processes. In OXMcultured cells, the most highly enriched GO term was ''triglyceride homeostasis'', and in XVIVOcultured cells the ''cholesterol biosynthetic process'' was highly enriched with 23/33 genes in this term upregulated in these cells . The term ''DNA replication initiation'' was enriched in XVIVO-cultured cells . To test whether XVIVO cells may therefore be in cell cycle, we measured total KI67, which transcriptionally has 2.66-fold higher expression in XVIVO-cultured cells, by immunostaining and flow cytometry. This confirmed expression is higher but note that only a small proportion of cells (XVIVO; 1.15% ± 0.90%, OXM; 0.54% ± 0.55%) are in cell cycle. In OXM, the inflammation-related term ''complement activation, classical pathway,'' cell adhesion, and several developmental terms are enriched. The enrichment of the cell adhesion term supports observations that OXM-cultured cells were more adherent than XVIVOcultured cells. Overall, when compared against adult microglia, both OXM-and XVIVO-cultured cells show similar but distinct expression profiles of genes within the GO terms mentioned above. However, OXM macrophages, in particular, display closer levels . Transcriptomic analysis has shown that, although cells cultured in these different media are highly similar, XVIVO-cultured macrophages display a greater tendency toward proliferation and an activated antiviral response, and OXM-cultured macrophages toward homeostasis, adhesion, and inflammation.
## Oxm-and xvivo-cultured macrophages have different classical and alternative activation profiles
To test whether the subtle transcriptomic differences between the cells differentiated in the two media resulted in functional differences, we measured cytokine secretion and polarization after stimulation. Cytokine secretion was measured at resting state and after classical (LPS and INF-g) or alternative (IL-4) activation, as described previously. Cytokine profiles under both conditions were consistent with the previous reports. OXM-cultured cells produced noticeably more GROa, ICAM-1, IL-1ra, and IL-8 in the resting state, while XVIVO-cultured cells produced more IP-10 (CXCL10). After classical activation, XVIVOcultured cells produced more tumor necrosis factor alpha (TNF-a), as well as low levels of IL-12p70, while OXM-cultured cells produced more RANTES (CCL5) and small quantities of IL-1b. Alternative activation suppressed secretion of resting state cytokines in both media. Consistent with the results above, LPS-induced TNF-a secretion from OXM-cultured cells quantified by enzyme-linked immunosorbent assay (ELISA) was lower than from XVIVO cells. Considering the differences in cytokine production, we next assessed macrophage surface marker polarization toward classically defined M1 and M2 states after activation. This was measured by immunostaining for CD86 and CD206, respectively. Although activation had no effect on CD45 expression in both conditions, CD86 expression signifi-cantly increased in OXM-cultured cells compared with XVIVO-cultured cells after classical activation (4.75-versus 0.81-fold). CD206 expression increased in both conditions after alternative activation and was not significantly different between conditions (1.86-versus 2.16-fold). In conclusion, although cells cultured under either condition can respond well to stimuli, they differ in their cytokine profile, and cells cultured in OXM display greater changes in typical macrophage polarization surface markers than those in XVIVO.
OXM-cultured macrophages are more susceptible to HIV-1 infection, but not to Zika virus Due to their positioning as sentinel cells throughout the body, tissue-resident macrophages are often the first members of the immune system to respond to infection. Consequently, several pathogens have evolved to use these cells as a host during early infection, including viruses, such as HIV-1 and Zika virus (ZIKV). In the case of HIV-1, macrophages are hypothesized to act as a potential reservoir for latent infection due to the long-lived, non-replicating state of these cells. To determine whether iPSC macrophages would be a suitable model for modeling this disease, macrophages were infected with a macrophage-tropic (CCR5-tropic Ba-L envelope), GFP expressing, HIV-1 pseudotype virus. Infectivity was measured by expression of GFP after 72 h by flow cytometry. A significantly higher proportion of macrophages cultured in OXM than in XVIVO were infected (6.76% ± 3.69% compared with 1.77% ± 1.23%;. Expression of the HIV-1 entry receptors CD4 and CXCR4 was not significantly different between conditions, and CCR5 was higher in XVIVO-cultured macrophages (mean flouresence intensity (MFI) relative to isotype: 6.16 ± 3.18 XVIVO versus 2.67 ± 1.52 OXM;. This difference in viral entry may be explained by the higher expression of known HIV-1 restriction factors in XVIVO-cultured cells, including the IFITM and the APOBEC families of proteins, and TRIM5a. To test whether differences in susceptibility are restricted to lentiviruses, we also assessed replication ZIKV strain in these cells. Cells produced in both media released virus after an eclipse phase of~6 h post infection with no difference in output titer ([3.36 ± 1.83] 310 7 pfu/mL XVIVO, [2.26 ± 1.14] 310 7 pfu/mL OXM at 48 h post infection;. Significance is shown when *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Therefore, OXM-cultured macrophages are a good model for HIV-1 infection, and differences in viral susceptibility between culturing conditions are not pan-viral.
# Discussion
The need to define an open-source medium for differentiating macrophages from iPSC was driven by two requirements. Firstly, to understand the finer changes in macrophage response to stimuli we must have a clear understanding of the resting conditions of these cells. Secondly, LC-MS analysis indicated the potential presence of several undisclosed molecules in the commercial serumfree alternative medium, XVIVO 15, some of which could inhibit, or polarize the macrophage response if present. Our analysis putatively identified isolariciresinol 9 0 -Oalpha-L-arabinofuranoside, a lignan glycoside, which is known to have anti-inflammatory effects, as well as methyl glucosinolate and dithionous acid. While the precise functions of these molecules are unknown, it is worth noting that indole glucosinolates have been reported to have potent anti-inflammatory effects. In its ionized form (dithionite), dithionous acid is a strong reducing agent capable of affecting the redox state of heme-containing enzymes, such as cytochrome b, which may affect cell physiology. We found that basing our new medium, OXM, on aD-MEM/F-12, did not prevent differentiation of macrophages from iPSCs. However, we did see considerable differences in both yield and morphology. OXM-cultured cells were significantly more adherent, with the increased expression of adherence-related genes. This appeared to limit the harvestability of the cells cultured in OXM to once per week, most probably due to inducing cell adherence to plastic seen when cultured in fresh differentiation medium, but does not explain the observed reduced yield of macrophages cultured in this medium. We have previously shown that XVIVO-cultured iPSC macrophages are significantly larger than blood MDMs, which are closer in size to OXM-cultured cells. Considering the very high levels of glucose in XVIVO medium, we suspected that there could be major differences in cell metabolism. We were therefore unsurprised to see that metabolism-related GO terms were among the most abundant terms differentially enriched between the media types. Whereas OXM-cultured cells were enriched for triglyceride homeostasis, reverse cholesterol transport and peptide catabolic process, XVIVO-cultured cells upregulated many genes involved in lipid metabolism and cholesterol biosynthesis, such as fatty acid synthase (FASN), 7-dehydrocholesterol reductase (DHCR7), and acetyl-CoA acetyltransferase 2 (ACAT2). Macrophage metabolism differs substantially between activated states (M1 versus M2), with fatty acid synthesis upregulated in M1 cells and fatty acid oxidation upregulated in M2 cells. Unsurprisingly then, we observed that XVIVO-cultured cells had higher expression of markers of classical M1 (pro-inflammatory) polarization. It is possible that this activated state in XVIVO is responsible for their relative insensitivity to LPS and interferon g-induced changes in CD86 expression, unlike OXMcultured cells, which displayed significant changes in this marker, although this does not appear to impact the cytokine response by these cells. Conversely, OXM-cultured cells showed similarities to non-classical or inflammatory monocytes, with increased expression of CD16. This potential pro-inflammatory phenotype may therefore be linked to the constitutive secretion of some inflammatory cytokines, such as IL-6 and GROa, but its importance in cellular function is unclear. Nevertheless, this cytokine profile observed is consistent with our previous workin which iPSC-derived microglia released similar cytokines in monoculture that were reduced when co-cultured with neurons. This suggests that our tissue-resident macrophages likely receive anti-inflammatory signals from neighboring stromal cells as part of a regulatory feedback system. Overall, neither resting OXM-or XVIVO-cultured macrophages clearly fell into conventional phenotypes, such as ''classical,'' ''non-classical,'' ''M1-activated,'' or ''M2-activated.'' However, iPSC-derived macrophages have been shown to more closely resemble primitive macrophages, and therefore such categories, which are largely derived from work with adult blood MDMs cultured in serum-containing medium, are an imperfect system for the phenotypic characterization of these cells. Instead, primitive macrophages, such as our iPSCderived macrophages, ought to be considered more along a spectrum of activation states, subtly influenced by nutrient availability and extracellular matrix and cell-cell interactions in the various tissue environments they may occupy.
One surprising observation was that XVIVO-cultured cells appear to be transcriptionally primed for an antiviral response. We found higher expression of many interferon-inducible genes and constitutive secretion of the interferon-inducible cytokine IP-10 (CXCL10). This clearly translated into a reduced infectivity of HIV-1 pseudotype virus in these cells compared with OXM-cultured cells, despite the cells having higher expression of the surface receptor CCR5. However, it did not result in a difference in ZIKV infection of these cells. This is supported by the fact that OXM-cultured cells had higher expression of the inter-feron receptors IFNAR2 and IFNGR2, and therefore are likely capable of mounting a strong antiviral response. It is worth noting that, although expression of interferon genes was lower in OXM-cultured cells, it was not absent. These cells may therefore be a better model of pre-viral exposure macrophages.
To conclude, we describe the successful development of a defined, and open-source medium for culturing human iPSC-derived macrophages: OXM. Subtle differences in phenotype and function in macrophages cultured in a commercial alternative highlight the importance of transparency in culturing techniques and demonstrates how direct comparisons without validation between culturing methods may not always be appropriate. We hope that the open-source nature of this medium will enable it to be used as a foundation for future improved differentiation protocols not only for macrophages, but also other iPSCderived cell types.
## Experimental procedures ipsc lines
The derivation and characterization of the iPSC lines used in this study are described elsewhere: SFC840-03-03, SFC841-03-01, and SFC856-03-04. See supplemental information for more information.
iPSC culture iPSCs were cultured in their stated medium on Geltrex (Gibco, no. A1413201)-coated tissue culture dishes and passaged using TrypLE Express (Gibco, no. 12604013). For 24 h after plating, medium was supplemented with 10 mM Y-27632 (Abcam, ab120129). Cells were incubated at 37 C, 5% CO 2 . See supplemental information for more information.
## Macrophage differentiation
iPSCs were differentiated into macrophages using a protocol based on that described previously. The updated method is as follows. AggreWell 800 plates were prepared by addition of 0.5 mL Anti-Adherence Rinsing Solution (STEMCELL Technologies, no. 07010) and centrifugation at 3,000 3 g for 3 min to remove bubbles from the microwells. Rinse solution was then aspirated and replaced with 1 mL of 23 concentrated EB medium (13 EB medium: OXE8 medium supplemented with 50 ng/mL BMP4 (PeproTech, no. PHC9534), 50 ng/mL VEGF (PeproTech, no. PHC9394), and 20 ng/mL SCF supplemented with 10 mM Y-27632. iPSCs were resuspended by washing with PBS, incubating in TrypLE Express for 3-5 min at 37 C, 5% CO 2 , followed by gentle lifting in aDMEM/F-12 to achieve single-cell suspension. Cells were counted and pelleted by centrifugation at 400 3 g for 5 min. After centrifugation, cells were resuspended at 4 3 10 6 cells/mL in OXE8 supplemented with 10 mM Y-27632 and 1 mL added to the AggreWell. The AggreWell plate was then spun at 100 3 g for 3 min with no braking to encourage even distribution of cells across microwells. Cells were incubated for 4 days at 37 C, 5% CO 2 , with daily feeding of 75% medium change with EB medium, by aspiration of 1 mL by pipette, and gentle addition of 1 mL fresh medium twice to avoid disturbance to the microwells. After 4 days, EBs were lifted from the plate using a Pasteur pipette and passed over a 40 mm cell strainer to remove dead cells, before washing into a tissue culture plate with differentiation medium . EBs were divided evenly into 2 T175 flasks and topped up to 20 mL with differentiation medium. Differentiation cultures were incubated at 37 C, 5% CO 2 , with weekly feeding of an additional 10 mL until macrophage precursors (PreMac) cells started to be produced. After this point, PreMac cells were collected weekly and a minimum of equal volumes of media to the volume removed were replaced into the differentiation cultures. Each harvest involved a 25%-50% medium change. Harvested cells could be stored or transported on ice for up to 48 h without loss of viability, and were either used directly or plated in appropriately sized tissue culture plates for further culturing in XVIVO or OXM macrophage medium for a further 7 days, with a 50% medium change on day 4.
## Cell count, size, and viability measurements
Cell count, size, and viability were measured using the Nucleo-Counter NC-3000 (ChemoMetec) after staining for live/dead cells using Solution-13 AO-DAPI stain . See the supplemental information for more details.
## Cell adhesion
Cells were lifted by incubation with Accutase and remaining adherent cells stained with NucBlue for imaging and quantification of nuclei by fluorescent microscopy. See the supplemental information for more details.
## Flow cytometry
Cells were stained directly without fixation in fluorescence-activated cell sorting buffer (PBS supplemented to 1% FBS, 10 mg/mL human-IgG (Sigma, no. I8640-100MG), and 0.01% sodium azide) and compared with isotype controls with the same fluorophores from the same company. Fluorescence was measured using the BD LSRFortessa X-20 (BD Biosciences) and analyzed on FlowJo version 10. See the supplemental information for more details.
## Cytokine and chemokine release
Cytokine and chemokine secretion was assessed as described previouslyusing the Proteome Profiler Human Cytokine Array Kit (R&D Systems, no. ARY005B). TNF-a was quantified by ELISA (Invitrogen, 88-7346-88) according to manufacturer's instructions. See the supplemental information for more details.
## Phagocytosis
Phagocytosis was measured by uptake of zymosan A (S. cerevisiae) BioParticles Alexa Fluor 488 (Thermo Fisher Scientific, no. Z23373) after 30 min. Uptake was measured by measurement of fluorescence using the BD LSRFortessa X-20. See the supplemental information for more details.
## Hiv-1 infection assay
Seven-day differentiated macrophages were infected for 72 h with a GFP-expressing HIV-1 lentiviral vector pseudotyped with strain Ba-L envelope virus diluted in macrophage medium. Percentage infected cells was determined by measurement of fluorescence using the BD LSRFortessa X-20 flow cytometer. See the supplemental information for more details.
## Zikv infection assay
Macrophages were infected with 2 3 10 6 pfu of ZIKV isolate MR1766 for 4 h before replacing with fresh medium. Supernatant was collected at the stated time points and stored for later quantification of virus by plaque assaying on Vero-76 cells. See the supplemental information for more details.
## Rna-seq
RNA was extracted from 1 3 10 6 macrophages cultured in a 6-well plate, differentiated from PreMac cells harvested at weeks 8, 10, and 12 of differentiation, and sequenced by Novogene. Results were analyzed in-house.
## Lc-ms analysis of media
Samples of the media were prepared for LC-MS analysis as described previously. See the supplemental information for more details.
## Quantification of glucose concentration
Glucose concentration in the media was quantified using a modified protocol for the Glucose (HK) Assay Kit (Supelco, no. GAHK20-1KT). See the supplemental information for more details.
# Statistical analysis
All statistical analysis used is reported in the figure legends and was carried out using GraphPad Prism version 8.2.1 or R version 4.0. In all cases data were considered significant when p < 0.05.
## Data and code availability
RNA-seq data and analysis have been deposited with the NCBI Gene Expression Omnibus. Theaccession number for the data in this paper is GEO: GSE171313. See the supplemental information for more details.
## Supplemental information
Supplemental information can be found online at https://doi.org/ 10.1016/j.stemcr.2021.05.018.
# Author contributions
|
Spinal metastasis: narrative reviews of the current evidence and treatment modalities
The treatment for spinal metastasis has evolved significantly during the past decade. An advancement in systemic therapy has led to a prolonged overall survival in cancer patients, thus increasing the incidence of spinal metastasis. In addition, with the improved treatment armamentarium, the prediction of patient survival using traditional prognostic models may have limitations and these require the incorporation of some novel parameters to improve their prognostic accuracy. The development of minimally-invasive spinal procedures and minimal access surgical techniques have facilitated a quicker patient recovery and return to systemic treatment. These modern interventions help to alleviate pain and improve quality of life, even in candidates with a relatively short life expectancy. Radiotherapy may be considered in non-surgical candidates or as adjuvant therapy for improving local tumour control. Stereotactic radiosurgery has facilitated this even in radioresistant tumours and may even replace surgery in radiosensitive malignancies. This narrative review summarizes the current evidence leading to the paradigm shifts in the modern treatment of spinal metastasis.
# Introduction
Spinal metastases are a challenging oncological problem, affecting up to 70% of cancer patients. [bib_ref] Metastatic spine disease: epidemiology, pathophysiology, and evaluation of patients, Perrin [/bib_ref] Up to 20% of spinal metastases become symptomatic, leading to pain, neurological deficit and disruption of health-related quality of life. [bib_ref] Symptomatic spinal metastasis: A systematic literature review of the preoperative prognostic factors..., Nater [/bib_ref] The common primary malignancies that cause spinal metastases include breast, lung and prostate cancer, accounting for up to twothirds of all cases. [bib_ref] Prognostic factors associated with survival in patients with symptomatic spinal bone metastases:..., Bollen [/bib_ref] [bib_ref] A score to identify patients with metastatic spinal cord compression who may..., Rades [/bib_ref] The frequently affected spinal regions are the thoracic, lumbar and cervical. [bib_ref] Physiopathology of spine metastasis, Maccauro [/bib_ref] In recent years, the advancement in cancer pharmacotherapy, including chemotherapy, targeted therapy, hormonetargeting drugs and immunotherapy, has led to substantially improved patient survival in almost all cancer types. [bib_ref] Current Challenges in Cancer Treatment, Zugazagoitia [/bib_ref] [bib_ref] Systemic Therapy for Locally Advanced and Metastatic Non-Small Cell Lung Cancer: A..., Arbour [/bib_ref] [bib_ref] The Influence of the Evolution of First-Line Chemotherapy on Steadily Improving Survival..., Noonan [/bib_ref] [bib_ref] Cancer immunotherapy: broadening the scope of targetable tumours, Van Den Bulk [/bib_ref] This amplifies the magnitude of the problems caused by spinal metastases. Furthermore, the novel development of stereotactic radiation surgery and minimally-invasive surgical techniques has allowed for good treatment outcomes in spinal metastases compared with traditional surgery. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] [bib_ref] Freehand Minimally Invasive Pedicle Screw Fixation and Minimally Invasive Decompression for a..., Fan [/bib_ref] [bib_ref] Minimally Invasive versus Open Surgery for Spinal Metastasis: A Systematic Review and..., Hinojosa-Gonzalez [/bib_ref] This review article aims to summarize the current evidence regarding the management of spinal metastases.
# Article search methodology
The two authors performed literature searches using electronic databases, including MEDLINE (PubMed V R ), Embase V R (Ovid), Cochrane Central Register of Controlled Trials (CENTRAL) and Cochrane Database of Systematic Reviews (CDSR) on 15 December 2021. The literature search was performed using the following search terms in combination with Boolean "OR" and "AND" phrases: "spinal metastasis", "spinal metastasis surgery", "spinal cord compression", "spinal tumour", "corticosteroid", "spinal fracture", "instability", "bisphosphonate", "denosumab", "skeletal related events", "NOMS framework", "SINS score", "Tohukashi score", "Tomita score", "Bollen score", "life expectancy prognostication", "embolization", "vertebroplasty", "kyphoplasty", "minimally invasive surgery", "minimal access surgery", "hybrid therapy", "stereotactic radiosurgery" and "complication". The resulting articles were screened for relevance to spinal metastasis. Articles published within the past 5 years, landmark articles and articles receiving high citations were included. The reference lists of the included studies were further reviewed for additional articles. Data from the included studies were reviewed independently by the two authors.
## Clinical presentation
The presenting signs and symptoms of spinal metastases depend on the tumour location and size. Up to 95% of patients experience pain as their initial symptom. [bib_ref] Metastatic spinal cord compression. Occurrence, symptoms, clinical presentations and prognosis in 398..., Bach [/bib_ref] [bib_ref] Prognostic factors in metastatic spinal cord compression: a prospective study using multivariate..., Helweg-Larsen [/bib_ref] Pain in spinal metastases can be caused by tumour invasion and expansion in the vertebral body causing pressure on the periosteum, spinal instability due to vertebral destruction, fractures or neural compression. In mechanically stable spinal metastases, the patient can develop pain during rest and severe pain that wake them up from sleep. By contrast, mechanical instability causes pain that is worsened by motion and improved with recumbency. Combined with vertebral collapses, this could lead to spinal deformity such as kyphosis. Nerve root compression or invasion by the tumour leads to pain, sensory and/or motor deficit following certain dermatomes. Bilateral radicular symptom, weakness of the limbs, gait abnormality, symmetric paresthesia, and bowel and bladdery dysfunction may suggest malignant spinal cord or cauda equina compression, requiring emergency treatment.
## Investigations
On radiographs, spinal metastases are osteolytic in 70%, osteoblastic in 8% and mixed osteolytic and osteoblastic in 21%. [bib_ref] Spinal metastases with neurological manifestations. Review of 600 cases, Constans [/bib_ref] Primary tumours that commonly cause osteolytic lesions include lung, breast, melanoma, thyroid and renal cancers. [bib_ref] Metastatic spine disease: epidemiology, pathophysiology, and evaluation of patients, Perrin [/bib_ref] Osteoblastic lesions can be caused by medulloblastoma, nasopharyngeal cancer, prostate cancer and urothelial cancer. [bib_ref] Metastatic spine disease: epidemiology, pathophysiology, and evaluation of patients, Perrin [/bib_ref] Mixed osteolytic and osteoblastic lesions can result from cervix, lung, breast and ovarian cancers. [bib_ref] Metastatic spine disease: epidemiology, pathophysiology, and evaluation of patients, Perrin [/bib_ref] Magnetic resonance imaging (MRI) of the whole spine remains the imaging modality of choice for detecting spinal metastases. [bib_ref] Detection of vertebral metastases: comparison between MR imaging and bone scintigraphy, Algra [/bib_ref] [bib_ref] MRI of bone metastases, Vanel [/bib_ref] [bib_ref] Detection of osseous metastases of the spine: comparison of high resolution multi-detector-CT..., Buhmann Kirchhoff [/bib_ref] The recommended views of study include sagittal T1 and T2 weighted images of the whole spine and axial T2 weighted images of the affected spinal region. [bib_ref] Diagnosis of vertebral metastasis, epidural metastasis, and malignant spinal cord compression: are..., Kim [/bib_ref] [bib_ref] Reliability analysis of the epidural spinal cord compression scale, Bilsky [/bib_ref] Hypointense lesions on T1-weighted images suggest spinal metastases. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] On T2-weighted images, hypointense and hyperintense signals represent osteoblastic and osteolytic lesions, respectively. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Gadolinium contrast enhancement is not strictly required for the detection of spinal metastases but is essential in diagnosing leptomeningeal metastases. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Furthermore, gadolinium contrast can disrupt spinal metastasis detection on T1weighted images. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Thus, non-contrast enhanced T1-weight imaging should be obtained beforehand. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] The recovery of patient's functional status after treatment associates with the duration of symptoms and neurological compression. [bib_ref] Predicting Neurologic Recovery after Surgery in Patients with Deficits Secondary to MESCC:..., Laufer [/bib_ref] The Dutch national guideline for the clinical management of spinal metastases recommends obtaining immediate MRI in clinical spinal cord compression or bilateral radicular compression. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] In those with radicular compression causing unilateral dermatomal/myotomal deficit, MRI should be done within 48 hours. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] In patients with unilateral radicular pain and localized pain, MRI should be completed within 1 and 2 weeks, respectively. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] In instances where MRI is contraindicated or not available, computed tomography of the spine with intravenous or intrathecal contrast should be alternatively considered. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Other imaging modalities such as plain X-ray and bone scintigraphy are not suitable for early diagnosis as metastatic lesions only become apparent with advanced bony destruction or soft tissue involvement. Furthermore, in purely osteolytic lesions, such as multiple myeloma, bone scintigraphy may yield a falsenegative result as their tracers are absorbed only by osteoblasts. [bib_ref] Comparison of radionuclide bone scans and magnetic resonance imaging in detecting spinal..., Gosfield E 3rd [/bib_ref] Tissue histology remains the gold standard for diagnosing skeletal metastasis, including in the spine. Other than at vertebral metastatic lesions, biopsy can also be performed from an epidural mass or from masses at other sites that are more easily accessed. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] [bib_ref] Diagnostic algorithm, prognostic factors and surgical treatment of metastatic cancer diseases of..., Szendr} Oi [/bib_ref] In cases of unknown primary malignancy, urgent positron emission tomography-computed tomography of the chest and abdomen are recommended. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] The Dutch national guideline advises that tissue diagnosis should be performed within 24 hours and 72 hours in the setting of clinical and radiological spinal cord compression, respectively. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] After biopsy, treatment, including surgical decompression, stabilization and radiotherapy, can be started immediately, whenever indicated, without waiting for the histological results.
Despite the high incidence of spinal metastasis in cancer patients, the treating physician should also be aware of other possible causes of pain and neurodeficit. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Differential diagnosis for back or neck pain includes osteoporotic spinal fractures, spinal infections, spondylosis, herniated nucleus pulposus, and referred pain from other sites, including visceral organs. Other pathology leading to neurodeficit in a patient with cancer includes leptomeningeal metastasis, radiation myelopathy, meningitis, epidural abscess/haematoma, spinal lipomatosis and intramedullary metastasis.
## Corticosteroids
In patients with symptomatic spinal cord compression, corticosteroids can be administered to reduce the swelling of the spinal cord. The available evidence suggests that the optimal dosage of intravenous dexamethasone is a 10 mg bolus, followed by 16 mg daily in divided doses. [bib_ref] Metastatic Spinal Cord Compression and Steroid Treatment: A Systematic Review, Kumar [/bib_ref] After completing definitive treatment, corticosteroid should be weaned off rapidly. Studies indicate that steroid administration leads to an increased number of patients with preserved ambulatory function at 1 year after treatment. [bib_ref] Metastatic Spinal Cord Compression and Steroid Treatment: A Systematic Review, Kumar [/bib_ref] [bib_ref] Effect of high-dose dexamethasone in carcinomatous metastatic spinal cord compression treated with..., Sørensen [/bib_ref] Furthermore, steroids should be given within 12 hours of spinal cord compression onset. [bib_ref] Treatment of spinal epidural metastasis improves patient survival and functional state, Oo [/bib_ref] A delayed steroid administration is associated with a six-fold risk of remaining non-ambulatory after treatment. [bib_ref] Treatment of spinal epidural metastasis improves patient survival and functional state, Oo [/bib_ref] In cases of asymptomatic radiological spinal cord compression, omitting corticosteroid does not lead to a negative impact on the patient's ambulatory capacity after radiotherapy. [bib_ref] Metastatic Spinal Cord Compression and Steroid Treatment: A Systematic Review, Kumar [/bib_ref] After radiation, the patient can experience significant pain due to inflammation along the irradiated regions. These pain flares can also be effectively treated with corticosteroid. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] In the setting of an unknown primary malignancy, corticosteroids should be deferred until after tissue histology/microbiology samples have been obtained. Evidence suggests that the histological diagnosis of haematological malignancy can be obscured by corticosteroid treatment. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Furthermore, the administration of steroids could lead to detrimental results in cases of spinal infection.
The use of steroids in spinal cord compression is not without complications. Their administration has been reported to be associated with complications such as pneumonia, infected decubitus ulcer, urosepsis, gastrointestinal bleeding, duodenal ulcer and peritonitis. [bib_ref] Treatment of spinal epidural metastasis improves patient survival and functional state, Oo [/bib_ref] [bib_ref] Initial bolus of conventional versus high-dose dexamethasone in metastatic spinal cord compression, Vecht [/bib_ref] [bib_ref] High incidence of serious side effects of high-dose dexamethasone treatment in patients..., Heimdal [/bib_ref] [bib_ref] Perforation of the gastrointestinal tract in patients receiving steroids for neurologic disease, Fadul [/bib_ref] However, these adverse effects are reported in historical studies where high dexamethasone dosage of 100 mg intravenous bolus and followed by 96 mg daily were used. [bib_ref] Treatment of spinal epidural metastasis improves patient survival and functional state, Oo [/bib_ref] [bib_ref] Initial bolus of conventional versus high-dose dexamethasone in metastatic spinal cord compression, Vecht [/bib_ref] [bib_ref] High incidence of serious side effects of high-dose dexamethasone treatment in patients..., Heimdal [/bib_ref] [bib_ref] Perforation of the gastrointestinal tract in patients receiving steroids for neurologic disease, Fadul [/bib_ref] When using low-dose steroids, the rate of gastrointestinal bleeding is reported to be as low as 1.9%. [bib_ref] Metastatic Spinal Cord Compression and Steroid Treatment: A Systematic Review, Kumar [/bib_ref] This may further be mitigated by the concurrent use of proton pump inhibitors. [bib_ref] Metastatic Spinal Cord Compression and Steroid Treatment: A Systematic Review, Kumar [/bib_ref] The risk of perioperative wound infection in low-dose steroid usage remains unreported. Overall, the benefits of steroid therapy on patient recovery and pain reduction are promising. Their usage should be weighed against the possible complications based on the presenting risk factors of complications in individual patients.
## Bisphosphonates and denosumab
Spinal metastasis patients can be affected by skeletal-related events (SREs), including bone pain, fracture, hypercalcaemia and spinal cord compression. [bib_ref] Diagnostic algorithm, prognostic factors and surgical treatment of metastatic cancer diseases of..., Szendr} Oi [/bib_ref] Bisphosphonates inhibit bone resorption mediated by osteoclasts which, in turn, leads to a lower risk of SREs in patients with selected primary cancer. [bib_ref] Diagnostic algorithm, prognostic factors and surgical treatment of metastatic cancer diseases of..., Szendr} Oi [/bib_ref] Zoledronic acid is the most commonly used intravenous bisphosphonate and it is approved for the treatment of bone metastasis in multiple myeloma and solid tumors. [bib_ref] Bone health in cancer patients: ESMO Clinical Practice Guidelines, Coleman [/bib_ref] Its administration effectively reduces serum calcium in patients with hypercalcaemia and delays the occurrence of SREs in patients with spinal metastases. [bib_ref] Bone health in cancer patients: ESMO Clinical Practice Guidelines, Coleman [/bib_ref] For the treatment of metastatic disease, 4 mg zoledronic acid is given intravenously every 3-4 weeks. [bib_ref] Bone health in cancer patients: ESMO Clinical Practice Guidelines, Coleman [/bib_ref] With this rate of frequency, the patient is faced with the risk of adverse events such as renal failure, hypocalcaemia and osteonecrosis of the jaw. [bib_ref] Bone health in cancer patients: ESMO Clinical Practice Guidelines, Coleman [/bib_ref] Thus, an alternative dosing interval of every 3 months has been proposed. Current evidence suggests that a 12-week dosing interval does not negatively affect SRE prevention. [bib_ref] A systematic review of dosing frequency with bone-targeted agents for patients with..., Hutton [/bib_ref] [bib_ref] Effect of Longer-Interval vs Standard Dosing of Zoledronic Acid on Skeletal Events..., Himelstein [/bib_ref] This longer interval has also been shown to improve patient compliance and reduce the occurrence of bisphosphonate-related adverse outcomes. [bib_ref] A systematic review of dosing frequency with bone-targeted agents for patients with..., Hutton [/bib_ref] [bib_ref] Effect of Longer-Interval vs Standard Dosing of Zoledronic Acid on Skeletal Events..., Himelstein [/bib_ref] Denosumab, a human monoclonal antibody that inhibits the receptor activator of nuclear factor kappa-B ligand, provides an alternative to zoledronic acid for SRE prevention. [bib_ref] Denosumab for the prevention of skeletal complications in metastatic castrationresistant prostate cancer:..., Smith [/bib_ref] [bib_ref] Denosumab compared with zoledronic acid for the treatment of bone metastases in..., Stopeck [/bib_ref] [bib_ref] Superiority of denosumab to zoledronic acid for prevention of skeletal-related events: a..., Lipton [/bib_ref] [bib_ref] Zoledronic Acid Versus Denosumab for Prevention of Spinal Cord Compression in Advanced..., Farii [/bib_ref] It is administered at a dose of 120 mg subcutaneously every 4 weeks. [bib_ref] Denosumab for the prevention of skeletal complications in metastatic castrationresistant prostate cancer:..., Smith [/bib_ref] Compared with zoledronic acid, denosumab is superior in delaying the occurrence of first and subsequent SREs in solid tumours and multiple myeloma. [bib_ref] Denosumab compared with zoledronic acid for the treatment of bone metastases in..., Stopeck [/bib_ref] [bib_ref] Superiority of denosumab to zoledronic acid for prevention of skeletal-related events: a..., Lipton [/bib_ref] Denosumab has shown no superiority in reducing the incidence of spinal cord compression compared with zoledronic acid. [bib_ref] Zoledronic Acid Versus Denosumab for Prevention of Spinal Cord Compression in Advanced..., Farii [/bib_ref] Denosumab also demonstrated a lower incidence of renal toxicity and acute phase reactions such as fever. [bib_ref] Denosumab for the prevention of skeletal complications in metastatic castrationresistant prostate cancer:..., Smith [/bib_ref] However, denosumab treatment is associated with a higher risk of hypocalcaemia and osteonecrosis of the jaw. [bib_ref] Denosumab for the prevention of skeletal complications in metastatic castrationresistant prostate cancer:..., Smith [/bib_ref] Unlike zoledronic acid, there is currently no evidence to support the length of the dosing interval of denosumab as it does not accumulate in the bone. Thus, a reduction in the dose frequency might negatively impact on the benefits of denosumab.
## Treatment considerations
The treatment of spinal metastases is palliative in nature. The aim is to alleviate pain, improve/maintain functional and neurological status and achieve local control of the tumour. [bib_ref] Surgery improves pain, function and quality of life in patients with spinal..., Quan [/bib_ref] [bib_ref] Survival and Clinical Outcomes in Surgically Treated Patients With Metastatic Epidural Spinal..., Fehlings [/bib_ref] Care of a patient with spinal metastasis involves a multidisciplinary team, including the initial treating specialist for the primary cancer, medical oncologist, radiation oncologist, radiologist and spine surgeon. The authors propose a treatment algorithm for spinal metastasis as presented in [fig_ref] Figure 1: Treatment algorithm for spinal metastasis [/fig_ref].
The presentation of spinal metastasis varies with tumour size, location and progression. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Surgical indications usually include intractable pain, mechanical instability and neurological compromise. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] The urgency of treatment varies according to the severity of the patient's symptoms. [bib_ref] Predicting Neurologic Recovery after Surgery in Patients with Deficits Secondary to MESCC:..., Laufer [/bib_ref] The Dutch national guideline recommends definitive treatment within 24 hours for symptomatic spinal cord compression, within 72 hours for asymptomatic radiological spinal cord compression and within 14 days for patients presenting with pain only. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] In 2013, the NOMS framework for guiding the comprehensive assessment of spinal metastasis patients was proposed. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] The framework consists of four pillars including neurological, oncological, mechanical stability and systemic condition parameters. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] First, the neurological and oncological parameters are evaluated together in combination. The neurological considerations include clinical neurological status and radiographic severity of epidural spinal cord compression (ESCC). The patient is clinically assessed for evidence of radiculopathy or myelopathy. Clinical myelopathy is highly associated with a high grade ESCC on MRI. The severity of ESCC is classified according to a 6-point grading scale according to the Spine Oncology Study Group. [bib_ref] Reliability analysis of the epidural spinal cord compression scale, Bilsky [/bib_ref] Grades 0, 1a, 1b and 1c represent low-grade ESCC where the thecal sac may be impinged, but no visible spinal cord compression is seen. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] Grades 2 and 3 represent high-grade ESCC, showing radiographical evidence of spinal cord compression. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] For oncological considerations, the primary malignancy is classified according to the anticipated response to available therapies, especially to conventional external beam radiotherapy (cEBRT). [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] Tumours with high-to-moderate radiosensitivity include haematological malignancies such as multiple myeloma, lymphoma and plasmacytoma, and solid tumours including prostate, breast, ovarian and neuroendocrine carcinomas. [bib_ref] A prospective evaluation of two radiotherapy schedules with 10 versus 20 fractions..., Rades [/bib_ref] [bib_ref] Final results of a prospective study comparing the local control of short-course..., Rades [/bib_ref] The majority of other solid tumours, including colon, non-small cell lung carcinoma, thyroid, renal cell carcinoma, melanoma, hepatocellular carcinoma, and sarcoma, are radioresistant to cEBRT. [bib_ref] A prospective evaluation of two radiotherapy schedules with 10 versus 20 fractions..., Rades [/bib_ref] [bib_ref] Final results of a prospective study comparing the local control of short-course..., Rades [/bib_ref] The patient without evidence of myelopathy or high-grade ESCC can be treated, without surgery, by cEBRT in radiosensitive cancer and by spinal stereotactic radiosurgery (SRS) in radioresistant tumours. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] In those presenting with highgrade ESCC and/or myelopathy in radiosensitive cancer, cEBRT can be administered as definitive treatment as rapid tumour response is anticipated. However, in the real-world setting, this patient group may be offered upfront decompressive spinal surgery in order to optimize their neurological outcome. In radioresistant tumours presenting with high-grade ESCC and/or myelopathy, spinal stabilization and decompression followed by spinal SRS are recommended. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] Mechanical stability of the metastatic spinal segment is evaluated according to the Spinal Instability Neoplastic Score (SINS), a score proven to have substantial to excellent interobserver reliability. [bib_ref] Intra-and interobserver reliability of the Spinal Instability Neoplastic Score system for instability..., Pennington [/bib_ref] [bib_ref] Spinal Instability Neoplastic Score (SINS): Reliability Among Spine Fellows and Resident Physicians..., Fox [/bib_ref] [bib_ref] Spinal instability neoplastic score: an analysis of reliability and validity from the..., Fourney [/bib_ref] Six parameters are evaluated in determining the stability of the metastatic spinal column: location, pain, alignment, type of bone lesion, radiographical spinal alignment, vertebral body collapse and posterolateral involvement of spinal elements. [bib_ref] A novel classification system for spinal instability in neoplastic disease: an evidence-based..., Fisher [/bib_ref] SINS scores of 0-6, 7-12 and 13-18 indicate stable, indeterminate stability and unstable spine, respectively. [bib_ref] A novel classification system for spinal instability in neoplastic disease: an evidence-based..., Fisher [/bib_ref] Indeterminate and unstable spines signify the need for consultation to a spine surgeon. Mechanical instability independently provides a surgical indication irrespective of ESCC severity and the presence of myelopathy. The necessity and plan of surgical treatment are decided based on the spine surgeon's experience and available facilities.
Lastly, the patient's condition is evaluated systemically to determine if the patient could withstand the formulated plan of treatment. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] In this process, the patient's comorbidities, life expectancy and disease burden are weighed for the risk and benefit of the treatment. With a palliative nature, spinal metastasis surgery aims to alleviate pain, provide stability, improve/maintain neurological function and facilitate an early return to systemic treatment.
## Survival prediction
The estimation of the patient's life expectancy by the treating physician can be inaccurate, especially in terminally ill patients. [bib_ref] How accurate are physicians' clinical predictions of survival and the available prognostic..., Chow [/bib_ref] [bib_ref] Predictive value of seven preoperative prognostic scoring systems for spinal metastases, Leithner [/bib_ref] [bib_ref] A systematic review of physicians' survival predictions in terminally ill cancer patients, Glare [/bib_ref] Thus, several predictive models of patient survival have been developed to prevent undertreatment or overtreatment.
Revised Tokuhashi, Tomita and modified Bauer scores are among the prognostic models widely used to predict remaining life expectancy. [bib_ref] Predictive value of seven preoperative prognostic scoring systems for spinal metastases, Leithner [/bib_ref] However, these scores were developed in the 1990s, prior to the recent advancements in pharmacotherapy that has significantly improved the outcomes of cancer treatment. These novel developments, including immunotherapy, targeted therapy, hormonal therapy and anti-vascular endothelial growth factor, have resulted in prolonged cancer progression-free survival and overall survival in virtually all types of malignancy. [bib_ref] Current Challenges in Cancer Treatment, Zugazagoitia [/bib_ref] [bib_ref] Systemic Therapy for Locally Advanced and Metastatic Non-Small Cell Lung Cancer: A..., Arbour [/bib_ref] [bib_ref] The Influence of the Evolution of First-Line Chemotherapy on Steadily Improving Survival..., Noonan [/bib_ref] [bib_ref] Cancer immunotherapy: broadening the scope of targetable tumours, Van Den Bulk [/bib_ref] A French nationwide retrospective study of 739 patients surgically treated between 2014-2017 demonstrated poor sensitivity and specificity of the traditional survival predictive scores. [bib_ref] Are spine metastasis survival scoring systems outdated and do they underestimate life..., Tabourel [/bib_ref] In this study, the accuracy of the revised Tokuhashi and Tomita scores were 42.8% and 25.6% in predicting survival of cancer patients in the modern era, respectively. [bib_ref] Are spine metastasis survival scoring systems outdated and do they underestimate life..., Tabourel [/bib_ref] Thus, the development of novel prognostic models has shifted toward incorporating individualized risk parameters for patients rather than using traditional risk score tables that predict survival from a generalized cluster of risk factors. The Skeletal Oncology Research Group (SORG) nomogram is a model composed of age, primary tumour type, Eastern Cooperative Oncology Group (ECOG) performance scale, presence of brain/visceral metastasis, number of spinal metastases, laboratory markers (white blood cell count and haemoglobin) and previous systemic treatment. [bib_ref] The SORG nomogram accurately predicts 3-and 12-months survival for operable spine metastatic..., Pereira [/bib_ref] The risk magnitude of each factor is weighted from the value individually measured in the specific evaluated patient. The SORG nomogram has demonstrated great accuracy in estimating survival at 3 months and 12 months for patients with operable spinal metastasis. [bib_ref] The SORG nomogram accurately predicts 3-and 12-months survival for operable spine metastatic..., Pereira [/bib_ref] Its accuracy in predicting 3-, 6-and 12-month survival was 90%, 71% and 78%, respectively. [bib_ref] The SORG nomogram accurately predicts 3-and 12-months survival for operable spine metastatic..., Pereira [/bib_ref] It is also amongst the limited scores showing good discriminative ability, consistently displaying an area under curve above 0.70 in receiver operating characteristic analysis for various time frames. [bib_ref] Are spine metastasis survival scoring systems outdated and do they underestimate life..., Tabourel [/bib_ref] [bib_ref] Prognostic scores for survival as decisional support for surgery in spinal metastases:..., Smeijers [/bib_ref] Novel survival prognostic models should strongly consider incorporating biological parameters such as targetable genetic mutations as treatment specifically targeting these mutations has been shown to significantly improve progression-free and overall survival in selected malignancies. [bib_ref] Are spine metastasis survival scoring systems outdated and do they underestimate life..., Tabourel [/bib_ref] The example of actionable mutations includes epidermal growth factor receptor/anaplastic lymphoma kinase in nonsmall-cell lung cancer, B-Raf in melanoma and hormonal status in breast cancer. [bib_ref] The function and therapeutic targeting of anaplastic lymphoma kinase (ALK) in non-small..., Golding [/bib_ref] [bib_ref] Targeted agents and immunotherapies: optimizing outcomes in melanoma, Luke [/bib_ref] [bib_ref] Breast Cancer: Current Molecular Therapeutic Targets and New Players, Nagini [/bib_ref] Despite the development of various survival prediction models, the treating physician should not strictly adhere to rigid predictions. Surgical management should be considered when indicated if postoperative systemic treatment is available.
## Indications for surgery
In the era of spinal stereotactic surgery, spinal surgery is indicated in radioresistant tumours presenting with high-grade ESCC and/or myelopathy. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] Mechanical instability also serves as an independent indication for surgical stabilization. [bib_ref] Spinal instability neoplastic score: an analysis of reliability and validity from the..., Fourney [/bib_ref] Surgery in spinal metastasis is known to be associated with a high complication rate up to 37%. [bib_ref] Survival, complications and outcome in 282 patients operated for neurological deficit due..., Jansson [/bib_ref] Thus, the risks and benefits of surgery should be cautiously weighed against the patient's expected survival and systemic condition. In patients with an expected survival of less than 3 months, surgery is generally not recommended. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Furthermore, the presence of more than three contiguous level spinal metastases precludes the surgeon from achieving stable spinal stabilization. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Thus, surgery should be avoided. In patients with life expectancy of more than 3 months, interventional procedures or minimal access surgery may be performed. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] Open spinal surgery requires a longer time to recover and is recommended in patients that are expected to live longer than 6 months. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] En bloc spondylectomy is a highly complex procedure and is only recommended in patients with an expected survival of at least 2 years. 63
## Non-surgical management: radiotherapy and stereotactic radiosurgery
Radiotherapy is the cornerstone of cancer treatment. The delivery of radiation destroys tumour cells by causing disruption of double-stranded DNA and damage to the tumour vasculature. [bib_ref] Ionizingradiation induced DNA double-strand breaks: a direct and indirect lighting up, Vignard [/bib_ref] However, organs adjacent to the target treatment area may also be collaterally affected, especially in cEBRT, leading to adverse effects such as oesophagitis, stomatitis and dermatitis. [bib_ref] Adverse Effects Of Radiation Therapy, Majeed [/bib_ref] Thus, SRS was developed, allowing image-guided delivery of a high radiation dose to a defined small target area with a sharp drop off gradient around the border of target radiation area. SRS delivers a relatively high radiation dose per fraction (>10 Gy) compared with cEBRT. [bib_ref] Radiobiological basis of SBRT and SRS, Song [/bib_ref] Such a high dosage leads to microvascular dysfunction and apoptosis, causing tumour hypoperfusion. Additionally, SRS helps generate a host immune response against the tumour cells, ultimately leading to tumour destruction with minimal damage to the adjacent tissues. [bib_ref] Radiobiological basis of SBRT and SRS, Song [/bib_ref] In non-surgical candidates that present with neurodeficit, a single fraction of 8 Gy cEBRT is advised in those with a short life expectancy. [bib_ref] 8Gy single-dose radiotherapy is effective in metastatic spinal cord compression: results of..., Maranzano [/bib_ref] [bib_ref] Single fraction radiotherapy is efficacious: a further analysis of the Dutch Bone..., Van Der Linden [/bib_ref] At least 30 Gy of cEBRT in divided fractions is recommended in patients that are expected to survive more than 6 months. In patients with spinal pain without spinal cord compression, a single fraction of 8 Gy cEBRT is recommended. [bib_ref] Evaluation of five radiation schedules and prognostic factors for metastatic spinal cord..., Rades [/bib_ref] [bib_ref] Preliminary results of spinal cord compression recurrence evaluation (score-1) study comparing short-course..., Rades [/bib_ref] Evidence suggests that a higher radiation dose does not provide a more effective pain reduction nor a longer duration of tumour response. [bib_ref] Single fraction radiotherapy is efficacious: a further analysis of the Dutch Bone..., Van Der Linden [/bib_ref] In patients with solitary metastasis or oligometastasis, a more aggressive plan of treatment, such as a combination of decompressive surgery and radiotherapy, is advised. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] This may help prolong progression-free survival and possibly lead to curative outcome in selected cases. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] In this setting, adjuvant radiotherapy is given at an ablative dose of 30-39 Gy in 10-13 divided fractions. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] More advanced radiation techniques such as SRS may also be considered. [bib_ref] Spinal metastases: Is stereotactic body radiation therapy supported by evidences?, Bari [/bib_ref] In radioresistant tumours, SRS is recommended over cEBRT, irrespective of the severity of ESCC. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] In radioresistant patients with low-grade ESCC, SRS can replace en bloc spondylectomy as a definitive therapy. [bib_ref] The NOMS framework: approach to the treatment of spinal metastatic tumors, Laufer [/bib_ref] This can be applied even in radioresistant tumours, showing a local control rate up to 88%. [bib_ref] Conventional Radiotherapy and Stereotactic Radiosurgery in the Management of Metastatic Spine Disease, Zhang [/bib_ref] Due to the high radiation dosage, SRS in delivered in only 1-3 fractions compared with 10-20 fractions in cEBRT. This also leads to a better patient compliance to radiation treatment. [bib_ref] State of the Art Treatment of Spinal Metastatic Disease, Barzilai [/bib_ref] However, SRS also leads to a higher risk of vertebral compression fracture (VCF), reported in up to 36% of irradiated vertebral segments. [bib_ref] Analysis of the Factors Contributing to Vertebral Compression Fractures After Spine Stereotactic..., Boyce-Fappiano [/bib_ref] Evidence suggests that limiting the radiation dose below 16-18 Gy per fraction could help mitigate the risk of SRS-induced VCFs. [bib_ref] Analysis of the Factors Contributing to Vertebral Compression Fractures After Spine Stereotactic..., Boyce-Fappiano [/bib_ref] In surgical candidates with a history of preoperative radiation, the rate of wound complications is reported as 6%. [bib_ref] No association between excessive wound complications and preoperative high-dose, hypofractionated, image-guided radiation..., Keam [/bib_ref] However, preoperative radiation, regardless of radiation dosage, was not shown to be an independent risk factor for wound problems. [bib_ref] No association between excessive wound complications and preoperative high-dose, hypofractionated, image-guided radiation..., Keam [/bib_ref] The evidence supporting the delay in adjuvant radiation after spinal metastasis surgery remains debatable. A previous study suggested that adjuvant SRS can be administered within 24 hours of surgical stabilization without the occurrence of wound complications at 90 days after surgery. [bib_ref] Stereotactic Radiotherapy Followed by Surgical Stabilization Within 24 h for Unstable Spinal..., Versteeg [/bib_ref] For cEBRT, a delay of at least 5-21 days after surgery is recommended before initiating radiotherapy in order to mitigate wound complications. [bib_ref] Risk factors for wound-related complications after surgical stabilization of spinal metastases with..., Jarvers [/bib_ref] [bib_ref] Timing of surgery and radiotherapy in the management of metastatic spine disease:..., Itshayek [/bib_ref] A delay of greater than 4 weeks has shown no benefit regarding wound problems. [bib_ref] Timing of Adjuvant Radiation Therapy and Risk of Wound-Related Complications Among Patients..., Azad [/bib_ref] Conversely, a 1-month delay increases local radiographic progression of spinal metastasis, leading to poorer quality of life, local control and overall survival. [bib_ref] Delayed postoperative radiotherapy increases the incidence of radiographic local tumor progression before..., Gong [/bib_ref]
## Preoperative embolization
Spinal metastasis surgery is complicated by extensive bleeding during surgery, especially in hypervascular primary tumours such thyroid and renal cell cancers. Massive intraoperative haemorrhage may cause surgical challenges by impairing surgical visualization, prolonging operative time and increasing the blood transfusion rate. A systematic review and meta-analysis has shown that preoperative embolization in hypervascular tumours leads to less intraoperative blood loss, a lower transfusion rate and a shorter operative time. [bib_ref] Effectiveness of Preoperative Embolization in Patients with Spinal Metastases: A Systematic Review..., Gao [/bib_ref] However, in nonhypervascular and mixed tumours, no significant differences were observed regarding transfusion requirement, intraoperative blood loss and operative time. [bib_ref] Effectiveness of Preoperative Embolization in Patients with Spinal Metastases: A Systematic Review..., Gao [/bib_ref] The complication rate and patient's overall survival were shown to be similar in the embolized and non-embolized groups. [bib_ref] Effectiveness of Preoperative Embolization in Patients with Spinal Metastases: A Systematic Review..., Gao [/bib_ref]
## Surgical management
## Interventional procedures
Vertebroplasty and kyphoplasty involve a polymethyl methacrylate (PMMA) bone cement injection into the vertebral body under image guidance. Cement augmentations provide spinal stability and thermal necrosis of the tumour leading to a substantial reduction in pain. [bib_ref] Controversial issues in kyphoplasty and vertebroplasty in malignant vertebral fractures, Papanastassiou [/bib_ref] [bib_ref] Efficacy of percutaneous vertebroplasty in patients with painful vertebral metastases: A retrospective..., Xie [/bib_ref] These procedures are indicated in patients with intractable pain without frank mechanical instability nor spinal cord compression.In cases where pain is caused by nerve root compression, vertebroplasty and kyphoplasty are not recommended as they do not lead to a reduction in tumour size. Studies have shown that these procedures reduce pain by 56-100% and lead to a complete response in 31% of patients. [bib_ref] Safety and efficacy of percutaneous vertebroplasty in malignancy: a systematic review, Chew [/bib_ref] [bib_ref] Osteoblastic and mixed spinal metastases: evaluation of the analgesic efficacy of percutaneous..., Calmels [/bib_ref] [bib_ref] Clinical and radiographic results of balloon kyphoplasty for treatment of vertebral body..., Dalbayrak [/bib_ref] [bib_ref] Palliation of compression fractures in cancer patients by vertebral augmentation: a retrospective..., Jha [/bib_ref] However, vertebroplasty and kyphoplasty are complicated by perioperative cement leakage. The incidence is reported in up to 75% of cases, with most being asymptomatic. [bib_ref] Prevention of cardiopulmonary embolization of polymethylmethacrylate cement fragment after kyphoplasty with insertion..., Agko [/bib_ref] Symptomatic cement leakage may cause catastrophic neural compromise, requiring emergency evacuation. [bib_ref] Risk Factors for Cement Leakage After Vertebroplasty or Kyphoplasty: A Meta-Analysis of..., Zhan [/bib_ref] The risk factors of cement leakage include greater cortical destruction, larger cement quantity and the use of low viscosity PMMA. [bib_ref] Risk Factors for Cement Leakage After Vertebroplasty or Kyphoplasty: A Meta-Analysis of..., Zhan [/bib_ref] Radiofrequency ablation (RFA) is a percutaneous procedure where high frequency radio waves are delivered into the metastatic vertebral body, where they induce a high temperature at the target site leading to tumour cell necrosis. RFA is often combined with vertebroplasty or kyphoplasty, showing effectiveness in reducing pain and disability up to 3 months. 91
## Decompression-alone surgery
Spinal decompression is the workhorse treatment for patients presenting with neurodeficit. However, decompression without stabilization is rarely indicated in the setting of spinal metastasis. [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] [bib_ref] State of the Art Treatment of Spinal Metastatic Disease, Barzilai [/bib_ref] Spinal invasion by tumours often involves the posterolateral elements of the vertebral column. Laminectomy alone poses a high risk of creating iatrogenic instability, possibly leading to postoperative morbidities.
## Minimal access surgery
To minimize blood loss and hasten recovery after surgery, minimal access surgical techniques have been developed over recent years. Spinal stabilization can be achieved with image-guided percutaneous instrumentation or the mini-open Wiltse approach, allowing for less soft tissue trauma. [bib_ref] Comparison of Minimally Invasive Tubular Surgery with Conventional Surgery in the Treatment..., Cui [/bib_ref] [bib_ref] Percutaneous Endoscopic Interlaminar Decompression of Hypervascular Spinal Metastases, Unsal [/bib_ref] Furthermore, spinal decompression can be done with minimal exposure via the mini-open midline approach, the use of tubular or expandable retractors and endoscopy. [bib_ref] Freehand Minimally Invasive Pedicle Screw Fixation and Minimally Invasive Decompression for a..., Fan [/bib_ref] [bib_ref] Minimally Invasive versus Open Surgery for Spinal Metastasis: A Systematic Review and..., Hinojosa-Gonzalez [/bib_ref] [bib_ref] Comparison of Minimally Invasive Tubular Surgery with Conventional Surgery in the Treatment..., Cui [/bib_ref] [bib_ref] Percutaneous Endoscopic Interlaminar Decompression of Hypervascular Spinal Metastases, Unsal [/bib_ref] Minimal access surgery has been shown to be associated with less perioperative complications, intraoperative blood loss, blood transfusion rate, duration of postoperative bed rest and length of hospital stay compared with open surgery. [bib_ref] Freehand Minimally Invasive Pedicle Screw Fixation and Minimally Invasive Decompression for a..., Fan [/bib_ref] [bib_ref] Minimally Invasive versus Open Surgery for Spinal Metastasis: A Systematic Review and..., Hinojosa-Gonzalez [/bib_ref] Comparable changes in postoperative pain, neurological outcome, ECOG performance score, odds of survival and the rate of reoperation have been demonstrated in both open and minimal access surgery. [bib_ref] Freehand Minimally Invasive Pedicle Screw Fixation and Minimally Invasive Decompression for a..., Fan [/bib_ref] [bib_ref] Minimally Invasive versus Open Surgery for Spinal Metastasis: A Systematic Review and..., Hinojosa-Gonzalez [/bib_ref] Prompt recovery from minimallyinvasive surgery allow cancer patients to quickly return to their oncological treatment. In open surgery, studies suggest a delay of 14 days to 1 month after surgery before initiating adjuvant systemic treatment such as chemotherapy, radiotherapy and targeted therapies in order to minimize the rate of wound complications. [bib_ref] Minimally Invasive versus Open Surgery for Spinal Metastasis: A Systematic Review and..., Hinojosa-Gonzalez [/bib_ref] [bib_ref] Clinical management of spinal metastases -The Dutch national guideline, Bollen [/bib_ref] In patients that had undergone minimal access surgery, cEBRT may be started 1 week postoperatively and SRS may be initially immediately after the surgery. [bib_ref] State of the Art Treatment of Spinal Metastatic Disease, Barzilai [/bib_ref] [bib_ref] Minimal access versus open spinal surgery in treating painful spine metastasis: a..., Yang [/bib_ref] [bib_ref] Management of radiated reoperative wounds of the cervicothoracic spine: the role of..., Disa [/bib_ref] Hybrid therapy
In the era of SRS, a hybrid therapy, combining separation surgery and postoperative SRS may be performed in radioresistant tumours. The posterior element of the metastatic spinal segment is removed unilaterally or bilaterally. The posterior longitudinal ligament is then resected. A plane is created between the tumour at the posterior vertebral body and the thecal sac, with a minimum space of 3 mm, allowing for the safe delivery of postoperative SRS. [bib_ref] 10-Year Trends in the Surgical Management of Patients with Spinal Metastases: A..., Orenday-Barraza [/bib_ref] This technique can be achieved via both minimal access and open spinal surgery. Evidence suggests that hybrid therapy yields an excellent outcome in local tumour control, with a recurrence rate of 4.3 to 22%. [bib_ref] 10-Year Trends in the Surgical Management of Patients with Spinal Metastases: A..., Orenday-Barraza [/bib_ref] The overall 1-year survival after surgery is reported to be up to 78.4%. [bib_ref] 10-Year Trends in the Surgical Management of Patients with Spinal Metastases: A..., Orenday-Barraza [/bib_ref] The rate of complications and reoperations are reported as 5.4-14% and 5%, respectively. 96
## En bloc spinal surgery
En bloc spinal surgery has reduced in popularity due to high perioperative morbidity and surgical complexity and has been replaced by hybrid therapy. Limited surgical indications remain in patients with an expected survival over 2 years, controlled primary tumour without extraspinal metastases, adequate cardiopulmonary reserve to tolerate surgery, acceptable preoperative performance status (ECOG 0-2) and in patients treated in centres without SRS in metastasis with known resistance to cEBRT. [bib_ref] Surgical Metastasectomy in the Spine: A Review Article, Kato [/bib_ref] The ideal candidate for en bloc surgery is a patient with a solitary metastasis that can be completely resected. It is the surgery of choice especially in single metastatic lesions from renal cell carcinoma and thyroid cancer. [bib_ref] Surgical Metastasectomy in the Spine: A Review Article, Kato [/bib_ref] The current guidelines for the management of isolated skeletal metastasis, including the spine, suggest complete metastatectomy when this is feasible in these two primary cancers, due to their poor response to radiation and systemic treatments relative to metastases from other tumour types. [bib_ref] Diagnostic algorithm, prognostic factors and surgical treatment of metastatic cancer diseases of..., Szendr} Oi [/bib_ref] [bib_ref] Surgical Metastasectomy in the Spine: A Review Article, Kato [/bib_ref] En bloc surgery may also be considered in spinal oligometastasis (3 contiguous vertebrae involvement) that cannot be effectively treated with other systemic or local modalities. [bib_ref] Surgical Metastasectomy in the Spine: A Review Article, Kato [/bib_ref] Overall, en bloc surgery can be performed via piecemeal total excision in the cervical spine and total en bloc spondylectomy in the thoracic or lumbar spine. The result of en bloc spinal surgery is reported to be excellent, with a local recurrence rate ranging from 1.1% to 15.3%. [bib_ref] Surgical Metastasectomy in the Spine: A Review Article, Kato [/bib_ref] In cases where complete tumour resection is achieved, superior local control and postoperative survival are expected when compared with cytoreductive surgery, including debulking of the tumour at the anterior vertebral column and decompression of the posterior elements. [bib_ref] Surgical Metastasectomy in the Spine: A Review Article, Kato [/bib_ref] However, the rate of major perioperative complications is up to 39.7% and the rate of perioperative mortality ranges from 1.3% to 9.7%. [bib_ref] Surgical Metastasectomy in the Spine: A Review Article, Kato [/bib_ref] Thus, it is recommended that en bloc spinal surgery should only be performed in high volume and experienced centres.
## Complications
Due to the frailty of cancer patients, spinal metastasis surgery is reported to have a high complication rate, ranging from 5.3% to 51%. [bib_ref] Complications and reoperations after surgery for 647 patients with spine metastatic disease, Pereira [/bib_ref] A previous study suggests that 30-day complications after spinal metastasis surgery leads to a worsened patient survival. [bib_ref] Complications and reoperations after surgery for 647 patients with spine metastatic disease, Pereira [/bib_ref] The occurrence of 30-day complications can be predicted by using the Charlson co-morbidity index (CCI) score where the patient's age, comorbidities and primary cancer are considered. [bib_ref] Complications and reoperations after surgery for 647 patients with spine metastatic disease, Pereira [/bib_ref] A CCI score 2 is a robust predictor of perioperative 30-day complications. [bib_ref] Complications and reoperations after surgery for 647 patients with spine metastatic disease, Pereira [/bib_ref] Treating physicians must be alert to the need for aggressive monitoring, intensive care, and prevention of possible adverse outcomes in surgical candidates with a high risk of complications.
Hardware failure is another major postoperative concern in patients treated with spinal stabilization. Bone quality of the metastatic vertebrae may be hindered by tumour invasion, systemic cancer treatment and poor nutritional status. [bib_ref] Survival, fusion, and hardware failure after surgery for spinal metastatic disease, Yee [/bib_ref] While the benefit of spinal fusion is robust in preventing hardware failure in degenerative spine surgery, its necessity is still debatable in the setting of spinal metastasis. [bib_ref] Complications and reoperations after surgery for 647 patients with spine metastatic disease, Pereira [/bib_ref] [bib_ref] Survival, fusion, and hardware failure after surgery for spinal metastatic disease, Yee [/bib_ref] [bib_ref] Outcome analysis of surgery for symptomatic spinal metastases in long-term cancer survivors, Barzilai [/bib_ref] In metastatic patients, evidence suggests a low fusion rate at 1 year, with only 16.1% achieving complete fusion, after the surgery. [bib_ref] Survival, fusion, and hardware failure after surgery for spinal metastatic disease, Yee [/bib_ref] The rate of spinal fusion over a longer time period may be difficult to evaluate due to the high mortality rate. The rate of hardware failure requiring reoperation is reported to be 4.2% at 1 year and 12.5% at 2 years. [bib_ref] Complications and reoperations after surgery for 647 patients with spine metastatic disease, Pereira [/bib_ref] [bib_ref] Survival, fusion, and hardware failure after surgery for spinal metastatic disease, Yee [/bib_ref] [bib_ref] Outcome analysis of surgery for symptomatic spinal metastases in long-term cancer survivors, Barzilai [/bib_ref] The risk factors for hardware failure include a construction greater than six levels, increasing age and chest wall resection. [bib_ref] Survival, fusion, and hardware failure after surgery for spinal metastatic disease, Yee [/bib_ref] A strategy for the prevention of hardware failure has been proposed by using fenestrated pedicle screw and PMMA augmentation in the vertebral body. [bib_ref] Safety and efficacy of cement augmentation with fenestrated pedicle screws for tumorrelated..., Massaad [/bib_ref] [bib_ref] Pullout strength of pedicle screws with cement augmentation in severe osteoporosis: a..., Chen [/bib_ref] [bib_ref] Bone Cement-Augmented Percutaneous Screw Fixation for Malignant Spinal Metastases: Is It Feasible?, Kim [/bib_ref] This augmentation is shown to decrease the risk of screw pull out in both open and minimal access spinal surgery. [bib_ref] Safety and efficacy of cement augmentation with fenestrated pedicle screws for tumorrelated..., Massaad [/bib_ref] [bib_ref] Pullout strength of pedicle screws with cement augmentation in severe osteoporosis: a..., Chen [/bib_ref] [bib_ref] Bone Cement-Augmented Percutaneous Screw Fixation for Malignant Spinal Metastases: Is It Feasible?, Kim [/bib_ref] Conclusion Spinal metastasis and spinal cord compression greatly affect the patient's quality of life and survival. Major advancements in systemic cancer treatment have led to longer patient survival and better treatment outcomes. The development of SRS and minimally-invasive surgical techniques have allowed for excellent local tumour control and a quicker return to systemic treatment. Treating physicians should be aware of the new survival prognostic models and treatment armamentarium in order to optimize the results of spinal metastasis treatment.
[fig] Figure 1: Treatment algorithm for spinal metastasis. ESCC, epidural spinal cord compression; MRI, magnetic resonance imaging; PET-CT, positron emission tomography-computed tomography; cEBRT, conventional external beam radiotherapy; SRS, stereotactic radiosurgery. The colour version of this figure is available at: http://imr.sagepub.com. [/fig]
|
Silicone Wristbands as Personal Passive Samplers
Active-sampling approaches are commonly used for personal monitoring, but are limited by energy usage and data that may not represent an individual's exposure or bioavailable concentrations. Current passive techniques often involve extensive preparation, or are developed for only a small number of targeted compounds. In this work, we present a novel application for measuring bioavailable exposure with silicone wristbands as personal passive samplers. Laboratory methodology affecting precleaning, infusion, and extraction were developed from commercially available silicone, and chromatographic background interference was reduced after solvent cleanup with good extraction efficiency (>96%). After finalizing laboratory methods, 49 compounds were sequestered during an ambient deployment which encompassed a diverse set of compounds including polycyclic aromatic hydrocarbons (PAHs), consumer products, personal care products, pesticides, phthalates, and other industrial compounds ranging in log K ow from −0.07 (caffeine) to 9.49 (tris(2-ethylhexyl) phosphate). In two hot asphalt occupational settings, silicone personal samplers sequestered 25 PAHs during 8-and 40-h exposures, as well as 2 oxygenated-PAHs (benzofluorenone and fluorenone) suggesting temporal sensitivity over a single work day or week (p < 0.05, power =0.85). Additionally, the amount of PAH sequestered differed between worksites (p < 0.05, power = 0.99), suggesting spatial sensitivity using this novel application.
# ■ introduction
Whether through work-related exposure, or interactions with the ambient environment, people are exposed to a complex mixture of natural and man-made chemicals. Chemical exposure may occur through dermal, oral, or inhalation pathways, and compounds such as polycyclic aromatic hydrocarbons (PAHs), pesticides, dioxins, and polychlorinated biphenyls have been studied for decades. [bib_ref] New developments in Standard Reference Materials (SRMs) for environmental forensics, Poster [/bib_ref] However, linking mixed chemical exposures to health effects is often difficult given the diversity of compounds and often low levels of exposure. [bib_ref] Understanding the human health effects of chemical mixtures, Carpenter [/bib_ref] [bib_ref] Monitoring human occupational and environmental exposures to polycyclic aromatic compounds, Brandt [/bib_ref] Even in occupational circumstances where many chemicals of concern are identified, linking exposure to biological end points is challenging given the long latency of some diseases, the magnitude of the potential doses or interactions, and other confounders with exposure such as life behaviors and genetic variability. [bib_ref] Monitoring human occupational and environmental exposures to polycyclic aromatic compounds, Brandt [/bib_ref] [bib_ref] Bitumen fumes: Review of work on the potential risk to workers and..., Binet [/bib_ref] Because of this complexity, there is now a push to capture life-course environmental exposures from before birth onward recognized as the "exposome". [bib_ref] Exposure science and the exposome: An opportunity for coherence in the environmental..., Lioy [/bib_ref] To understand linkages between the exposome and resulting toxicity, researchers are developing new technologies and methods to characterize exposure to an ever larger range of compounds. Often however, environmental data are gathered from single time points which may not reflect average exposures in profile or magnitude. In contrast to single time point samples, active and passive sampling has been used to monitor PAHs and other organic chemicals with stationary and personal samplers. Although stationary samplers are used in several occupational studies, personal samplers have the advantage of being more relevant to an individual. [bib_ref] Occupational and indoor air exposure to persistent organic pollutants: A review of..., Bohlin [/bib_ref] The most common personal samplers are active devices that pump air through filters that are extracted for target compounds. However, active personal devices are relatively expensive, require energy, and ultimately limited to implement on a wide scale. [bib_ref] Occupational and indoor air exposure to persistent organic pollutants: A review of..., Bohlin [/bib_ref] An alternative to active monitoring is passive sampling.
A report of a personal passive sampler was first published in 1973, [bib_ref] Personal monitoring device for gaseous contaminants, Palmes [/bib_ref] but most growth in passive sampling and personal monitoring has occurred within the past decade. [bib_ref] Passive sampling and/or extraction techniques in environmental analysis: A review, Namiesńik [/bib_ref] Passive sampling devices (PSDs) are used to sequester organic molecules through passive diffusion from water or air, and provide time-weighted averages of chemical concentrations. [bib_ref] Passive sampling for longterm monitoring of organic pollutants in water, Kot [/bib_ref] Because chemicals continually accumulate in PSDs, the sensitivity of analytical detection is increased, and samplers represent time-averaged concentrations rather than episodic contamination. [bib_ref] Passive sampling for longterm monitoring of organic pollutants in water, Kot [/bib_ref] PSDs have been used for personal monitoring starting with water vapor and SO 2 measurements, [bib_ref] Personal monitoring device for gaseous contaminants, Palmes [/bib_ref] and have expanded to include organic contaminant classes like PAHs and PCBs in recent years. [bib_ref] Occupational and indoor air exposure to persistent organic pollutants: A review of..., Bohlin [/bib_ref] Materials used in passive sampling vary widely, and have included simple matrices like activated carbon,as well as complex polymers like polyethylene and silicone. [bib_ref] Passive sampling and/or extraction techniques in environmental analysis: A review, Namiesńik [/bib_ref] Historically, most personal monitoring samplers measure only one or a few compounds, [bib_ref] Passive dosimeters for nitrogen dioxide in personal/indoor air sampling: A review, Yu [/bib_ref] but recently the applications of PSDs have expanded to entire compound classes in order to assess chemical mixtures. One recent example includes polyurethane foam (PUF) used as personalized passive samplers. 14 While this material successfully sequesters hydrophobic PAHs, PCBs, and certain pesticides, [bib_ref] Occupational and indoor air exposure to persistent organic pollutants: A review of..., Bohlin [/bib_ref] it is unclear whether these samplers would be able to target less hydrophobic compounds and more volatile pesticides. [bib_ref] Pesticides in the atmosphere of the Mississippi River Valley, part IIAir, Foreman [/bib_ref] In addition, it is unclear if future work will be able to exclude nontarget particulate sizes with a protective surface. 14 In contrast, compounds reported in silicone PSDs represent only the vapor phase, which may encompass 34−86% of the toxicological dose of PAHs in industrial exposures. [bib_ref] Risk assessment related to atmospheric polycyclic aromatic hydrocarbons in gas and particle..., Ramirez [/bib_ref] We wanted to demonstrate a PSD that can be used to measure PAHs and volatile organic compounds, but also one that captures personal care products, pesticides, and other compounds of emerging concern with a wide range of physicochemistry. Silicone is known to absorb a wide range of compounds in field applications from volatile benzene, toluene, ethylbenzene, and xylene compounds to more hydrophilic contaminants containing hydroxyl, ketone, or carboxyl groups. [bib_ref] Applications of polydimethylsiloxane in analytical chemistry: A review, Seethapathy [/bib_ref] [bib_ref] Passive sampling for target and non-target analyses of moderately polar and nonpolar..., Allan [/bib_ref] We hypothesized that silicone material in commercial wristbands could be modified for use as a personal passive sampler in much of the same way silicone is used and demonstrated in environmental studies. By using wristbands as a personal passive sampler, it would have advantages as compared with active samplers mentioned previously, and result in data that represents time-weighted, vapor-phase concentrations. Our objectives were 3-fold: modify commercially available wristbands for analytical extraction, identification, and quantitation of target compounds; demonstrate sequestration of a wide physiochemical range to broaden potential usage of the personal PSD; and finally, present quantitated data in real-world occupational settings to examine if samplers provide useful sensitivity and selectivity in this novel application.
## ■ experimental section
Wristband and Precleaning Experiments. All solvents were Optima-grade (Fisher Scientific, Pittsburgh, PA) or equivalent, and all laboratory glassware or other tools were solvent-rinsed before use. Any water used in postdeployment cleaning or initial washes of commercial silicone was filtered through a Barnstead D7389 purifier (Dubuque, IA). Commercially available silicone bracelets were purchased in two sizes (width: 1.3 and 2.5 cm; inner diameter: 6.4 and 6.7 cm respectively; 24hourwristbands.com, Houston, TX), and were used in several configurations throughout the study [fig_ref] Figure 1: Examples of silicone personal sampling samplers [/fig_ref]. Weights of smaller width wristbands were similar regardless of pigmentation (orange: 5.67 ± 0.02 g; clear: 5.68 ± 0.02 g; orange/white: 5.71 ± 0.02 g; n = 15 for each color). Larger 2.5 cm wristbands weighed 10.38 ± 0.02 g, but only the smaller sized wristbands were used in quantitative work described below. Before deployment, oligomers and other material that might interfere with future chemical analyses were reduced with various solvents in material/solvent ratios similar to other published work. After several experiments to optimize the process with less solvents or cleaning time, the final procedure used nominally ≤65 g of silicone in 800 mL of mixed solvent for 5 exchanges. A mixture of ethyl acetate/ hexane (1:1, v:v) was used for the first three exchanges, and ethyl acetate/methanol (1:1, v:v) was used for the last two exchanges. Each exchange occurred after a minimum of 2.5 h at 60 rotations per minute (VWR orbital shaker, Radnor, PA). Afterward, solvent-cleaned wristbands were placed in stainless steel canisters (AEB Kegs, Delebio SO, Italy) and dried under PUF filtered vacuum (≤3 days). Dried wristbands throughout the study were stored in either amber glass jars or in polytetrafluoroethylene (PTFE) airtight bags (Figure 1b−c) at 4°C until needed. Prior to occupational field deployment, two wristbands from a batch of precleaned silicone were assessed to ensure cleaning processes were adequate for Configurations of wristbands used in the study including a "single" wristband, one cut and worn as a "lapel", and as a "stacked" wristband in which only the outer band was analyzed; (b−c) bags used for transport that were attached to track participant ID and exposure time in the occupational deployments; (d) single wristband deployment (debossed writing as pictured: "OSU EINOME", Oregon State University Environmental Integrated Organic Monitor of Exposure). quantitative analyses. Specifically, if the highest background peak had an area less than 15-fold of a spiked internal standard of 500 ng/mL, then that background level was considered adequate for deployment and quantitative analysis.
## Environmental science & technology
Extraction and Exposure Optimization. Reports of extraction of silicone vary widely from single soaking periods, to extended Soxhlet extraction over 90 h. [bib_ref] Passive sampling: Partition coefficients for a silicone rubber reference phase, Yates [/bib_ref] [bib_ref] Application of a silicone rubber passive sampling technique for monitoring PAHs and..., Monteyne [/bib_ref] To determine an adequate extraction method, precleaned silicone wristbands were infused with four deuterated PAHs similar to a previous method. [bib_ref] Spiking of performance reference compounds in low density polyethylene and silicone passive..., Booij [/bib_ref] Briefly, acenaphthylene-D8, fluorene-D10, phenanthrene-D10, and pyrene-D10 were pipetted into a 1 L jar filled with approximately 50−100 g of silicone and a methanol/water (1:1, v:v) solution. Compounds were allowed to equilibrate for three days since the ratio of methanol/water used was 1:1 rather than 4:1 as originally described. [bib_ref] Spiking of performance reference compounds in low density polyethylene and silicone passive..., Booij [/bib_ref] Using a 1:1 ratio requires less deuterated compounds in the infusing solution since more will partition to the silicone. Wristbands were dried as previously described, and then three rounds of extraction at two time periods of either 2 or 24 h were used to examine efficiency (Supporting Information (SI) [fig_ref] Figure 1: Examples of silicone personal sampling samplers [/fig_ref].
Postdeployment cleaning consisted of two rinses with purified water, and one rinse with isopropyl alcohol to reduce any water residue and further remove surface particulates (SI [fig_ref] Figure 2: Total ion chromatograms of wristband extracts through stages of cleaning and overspike... [/fig_ref]. Field samplers were extracted twice with 100 mL of ethyl acetate on an orbital shaker at 60 rotations per minute (VWR) for nominally 2 h each time. Both rounds of extraction were combined and reduced to 1 mL (measured with premarked glassware) with closed-cell evaporators (Biotage LLC, Charlotte, NC). Samples were transferred and stored in amber chromatography vials at 4°C.
To examine whether PAHs would degrade after sorption to the wristband, or if field/handling conditions would influence exposure concentrations, we again infused wristbands with several PAHs (fluorene-d10, benzo[b]fluoranthene-d12, fluorene, pyrene, and benzo[b]fluoranthene) and either exposed outdoors (in sun or shade) or within PTFE storage bags at approximately −20°C, 23 and 35°C. Additional details are described in the SI. Silicone PSDs were extracted and stored as described above.
Instrumental Analysis. Samples screened for 1182 chemicals of concern were analyzed using retention time locking automatic mass spectral deconvolution and identification software (AMDIS) on an Agilent 5975B gas chromatograph−mass spectrometer (GC-MS) with a DB-5MS column (Agilent) at an electron impact mode of 70 eV. The spectra were compared against in-house and purchased libraries of compounds that included pesticides, polychlorinated biphenyls (PCBs), parent and substituted PAHs, pharmaceuticals, phthalates, as well as other compounds. Prior to PAH and OPAH instrumental analyses, perylene-d12, and fluorofluorenone-C 13 were spiked at 500 ng mL −1 as internal standards, respectively. Instrument parameters to analyze PSD extracts for 33 PAH compounds and 22 OPAHs have been described previously. [bib_ref] Impact of the deepwater horizon oil spill on bioavailable polycyclic aromatic hydrocarbons..., Allan [/bib_ref] [bib_ref] An analytical investigation of 24 oxygenated-PAHs (OPAHs) using liquid and gas chromatography−mass..., O'connell [/bib_ref] Analyses were performed on the same GC-MS and column described above but in selective ion mode rather than full scan. In addition, deuterated homologues of PAHs [bib_ref] The use of active personal dosemeters as a personal monitoring device: Comparison..., Boziari [/bib_ref] and OPAHs (2) were used during the extraction process to monitor potential losses in the laboratory. [bib_ref] Impact of the deepwater horizon oil spill on bioavailable polycyclic aromatic hydrocarbons..., Allan [/bib_ref] [bib_ref] An analytical investigation of 24 oxygenated-PAHs (OPAHs) using liquid and gas chromatography−mass..., O'connell [/bib_ref] For PAHs, sample concentrations were determined by the relative response of deuterated surrogates to target analytes in a 9point calibration curve with correlation coefficients for each analyte greater than 0.98. OPAHs were also quantitated with a 9-point calibration curve with correlation coefficients >0.99, but were not recovery corrected due to the availability of appropriate surrogates. [bib_ref] An analytical investigation of 24 oxygenated-PAHs (OPAHs) using liquid and gas chromatography−mass..., O'connell [/bib_ref] Ambient Demonstration. To determine if silicone wristbands could sequester a wide range of organic compounds, a public query was made to collect volunteers. Participants were instructed to wear a wristband continuously for 30 days including bathing, sleeping, or other activities. A total of 30 precleaned and dried wristbands were placed inside three amber jars, and metal tongs were used by participants as they took one or two wristbands to wear. A sign-out sheet was used to track the number of wristbands a participant took (1 or 2), but no surnames or personal information was asked or collected during this initial demonstration. At the end of the 30 day period, small (250 mL) amber jars were used to collect each individual wristband and were stored at −20°C until postdeployment cleaning and extraction. In addition, three nondeployed wristbands were placed inside amber jars at room temperature to serve as controls for potential laboratory or processing contamination.
Occupational Application. To meet our final objective, we deployed silicone PSDs to roofers using hot asphalt since occupational environments represent relevant exposures, and we focused on PAH quantitation since this compound class is of toxicological concern for this occupation. [bib_ref] Monitoring human occupational and environmental exposures to polycyclic aromatic compounds, Brandt [/bib_ref] [bib_ref] Bitumen fumes: Review of work on the potential risk to workers and..., Binet [/bib_ref] Our occupational study was approved by the institutional review board (IRB) of Oregon State University, and roofers were recruited to wear the silicone personal samplers while working with hot asphalt. To see if reduced skin contact would improve chemical analyses, each roofer wore three designs of silicone personal samplers simultaneously: a single wristband like the initial ambient study, a cut wristband pinned as a lapel on a shirt collar, and a stacked wristband in which an inner silicone band protected the outer band from sweat and oils [fig_ref] Figure 1: Examples of silicone personal sampling samplers [/fig_ref]. Hereafter, each configuration will be referred to as either single, lapel, or stacked, respectively. In the first setting, three workers wore PSDs for both a single day (approximately 8 h), and for a representative workweek (32−39 h) while refurbishing a roof at an active worksite. Due to availability, only the single and lapel configurations were worn for 8 h while all three configurations were worn for the representative workweek. Both the single and multiday deployments began on the same day. At the second site, five preapprentice roofers wore all three silicone PSD configurations throughout an 8-h shift at a training facility. Before either deployment, each sampler was placed into prelabeled PTFE bags [fig_ref] Figure 1: Examples of silicone personal sampling samplers [/fig_ref]. Nitrile gloves were used before and after each shift by nonparticipants when handling PSDs. In the case of the multiday deployment, PSDs were returned at the beginning of the next available shift after overnight storage at 4°C. Travel blanks consisting of precleaned silicone PSDs in PTFE bags were used at each setting and type of deployment (single or multiday). Additional roofing information is available in SI.
Quality Control and Statistics. Over 40% of instrumental samples were for quality control (QC) purposes. QC samples not already mentioned included instrument check standards ran before and after each set of samples (every 10 or less) as well as laboratory solvent blanks. PAHs and OPAHS in check standards had to be within 20% of the true value before samples would be allowed to proceed with analysis. Nondeployed wristbands were used during postdeployment cleaning to ensure there was no contamination or compound carryover between samples. For AMDIS analysis, only compounds above a 60% mass spectral match were considered for chemist review. Deconvoluted results are compared against reference spectra for each target analyte, and if multiple lines of evidence (ex: correct ratios of ions, larger ions more representative of the parent ion, and retention time match) are present, then an analyte is considered as identified in the sample. Any compounds identified in controls or laboratory blanks were removed from the initial ambient demonstration since AMDIS results are descriptive as presented. Any quantitated compounds in blanks from PAH or OPAH methods are described in the Results section.
Multivariate statistics were performed on ambient data using R statistical software (R development core team, Vienna, Austria). Identification data was converted into binary values, and a nonmetric multidimensional scaling model was used to graphically represent the data with Jaccard distance. For occupational comparisons, after normality and equal variance tests passed criteria, parametric t tests were performed in Sigmaplot (Systat Software Inc., San Jose, CA) with an assumed alpha value of 0.05. The power and p-value for the t tests are listed for each result described below. In this demonstration, PAHs were not back-calculated to atmospheric concentrations since meaningful comparisons could already be made and address our original objectives of sensitivity and selectivity in a real-world exposure.
■ RESULTS AND DISCUSSION Laboratory Method Development. Initially, silicone background was reduced similar to other methods in either solvents used or extraction times. [bib_ref] Field performance of seven passive sampling devices for monitoring of hydrophobic substances, Allan [/bib_ref] [bib_ref] Using silicone passive samplers to detect polycyclic aromatic hydrocarbons from wildfires in..., Schafer [/bib_ref] [bib_ref] Polymer selection for passive sampling: A comparison of critical properties, Rusina [/bib_ref] [bib_ref] Spiking of performance reference compounds in low density polyethylene and silicone passive..., Booij [/bib_ref] While this initial methodology allowed compounds to be identified without postextraction silica cleanup used in other work, [bib_ref] Field performance of seven passive sampling devices for monitoring of hydrophobic substances, Allan [/bib_ref] [bib_ref] Spiking of performance reference compounds in low density polyethylene and silicone passive..., Booij [/bib_ref] [bib_ref] Application of a silicone rubber passive sampling technique for monitoring PAHs and..., Monteyne [/bib_ref] improvements were sought to further reduce cleanup time and siloxane background [fig_ref] Figure 2: Total ion chromatograms of wristband extracts through stages of cleaning and overspike... [/fig_ref]. After optimization experiments, cleanup was improved by incorporating hexanes in addition to ethyl acetate and methanol. Additionally, a reduction of precleaning time was achieved in under two days versus five [fig_ref] Figure 2: Total ion chromatograms of wristband extracts through stages of cleaning and overspike... [/fig_ref]. After the cleanup method was finalized, all compounds reported in this work (PAHs and OPAHs) were spiked with silicone wristbands, extracted through the laboratory procedure, and quantitated within 26% of the true value [fig_ref] Figure 2: Total ion chromatograms of wristband extracts through stages of cleaning and overspike... [/fig_ref].
In extraction efficiency experiments, over 90% of the total amount of acenaphthalene-D8, fluorene-D10, phenanthrene-D10, and pyrene-D10 were extracted with the first round of ethyl acetate (SI [fig_ref] Figure 1: Examples of silicone personal sampling samplers [/fig_ref]. Less than 6% and 5% were extracted with a second and third round, respectively. Variability of infused wristbands used for these extraction experiments had less than 13% relative standard deviation across all time points and compounds. PAHs with lower hydrophobicity had lower extraction efficiency over the first round of solvent, but all four compounds were ≥96% of the final extracted amount after two rounds of extraction. The total amount of compounds did not differ whether treatments were 2 or 24 h (892 ± 60 ng/mL or 878 ± 47 ng/mL, respectively).
In the sun/shade experiment, we did not observe any statistical difference between PAHs over a four-hour period (SI [fig_ref] Figure 4: Worksite PSDs [/fig_ref]. This preliminary evidence suggests there was no photodegradation of 5 PAHs once sequestered into the PSD, which is consistent with a previous observation that PAHs sorbed to fly ash have reduced photo-oxidation.Further study would be needed to examine PAH stability of longer time periods and varied irradiance, but for the purposes of this initial paper, potential degradation of sorbed analytes was not of concern. Additionally, no difference was observed among transport temperatures in PFTE bags (SI . Our data suggests that transportation in PTFE bags with temperatures as high as 35°C and transport times up to 72 h does not affect target analyte recovery. Stability during PFTE transport is consistent with similar work with PAHs and pesticides in polyethylene passive samplers (manuscript submitted). Ambient Demonstration. A wide range of compounds were identified from the ambient wristband extracts from 22 participants, with log K ow properties ranging from −0.07 (caffeine) to 9.49 (tris(2-ethylhexyl) phosphate) listed in [fig_ref] Table 1: Compounds Identified from AMDIS Spectra against Chemical Libraries during Ambient Exposures a [/fig_ref]. In contrast, a recent publication required that two PSDs materials together in an environmental deployment were needed to obtain a similar range of chemistry (K ow : caffeine −0.07 to DDT 6.91). [bib_ref] A holistic passive integrative sampling approach for assessing the presence and potential..., Petty [/bib_ref] In total, 49 different compounds were identified in our study, including PAHs, consumer and personal care products, pesticides, phthalates, and other industrial compounds [fig_ref] Table 1: Compounds Identified from AMDIS Spectra against Chemical Libraries during Ambient Exposures a [/fig_ref]. Most individual compounds were PAHs, or consisted of industrial compounds typically used as flame retardants, plasticizers, or used in synthetic material manufacturing [fig_ref] Table 1: Compounds Identified from AMDIS Spectra against Chemical Libraries during Ambient Exposures a [/fig_ref]. The two most detected compounds are diethyl phthalate (all samples) and tonalide (20 of 23), both of which are used in personal care products like fragrances or cosmetics. 29 Home-use pesticides like N,N-diethyl-m-toluamide (DEET) and fipronil (pet flea medicine) were identified in several samples as well as consumer product ingredients like caffeine and nicotine. Many of the individual compounds listed in [fig_ref] Table 1: Compounds Identified from AMDIS Spectra against Chemical Libraries during Ambient Exposures a [/fig_ref] have been previously sequestered in environmental studies using silicone, [bib_ref] Passive sampling for target and non-target analyses of moderately polar and nonpolar..., Allan [/bib_ref] [bib_ref] Occurrence of contaminants of emerging concern along the California coast (2009−10) using..., Alvarez [/bib_ref] and all of the compound classes have been associated with human exposures through previous research. [bib_ref] Understanding the human health effects of chemical mixtures, Carpenter [/bib_ref] [bib_ref] Current approaches to trace analysis of pharmaceuticals and personal care products in..., Buchberger [/bib_ref] [bib_ref] The use of biomonitoring data in exposure and human health risk assessments, Albertini [/bib_ref] Overall, results from these personal silicone samplers represent a wide diversity of bioavailable compounds, and appear to be different among individual participants using the nonmetric multidimensional scaling model (SI [fig_ref] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day [/fig_ref]. Further enhancements and separation of unique profiles of exposure should be possible once squalene and free fatty acids detected from full scan analysis are reduced by minimizing skin contact (in placement or duration as demonstrated in the occupational study).
Along with prominent skin components, caffeine and other relatively nonvolatile compounds in [fig_ref] Table 1: Compounds Identified from AMDIS Spectra against Chemical Libraries during Ambient Exposures a [/fig_ref] were likely taken up through direct contact, and wristband passive samplers may be beneficial in cases where less volatile metabolites or unchanged parent compounds are targeted in human exposure estimates. However, further evidence of this route of exposure is needed before studies may exploit this potential sampling attribute while separating out interferences from skin components. For the purposes of the occupational study discussed below, several silicone configurations (single, lapel, and stacked) were used to evaluate changes, if any, with skin contact during these shorter exposure periods.
Occupational PAH Results. A total of eight roofers wore silicone passive samplers, with three at an active worksite (nos. 1−3, [fig_ref] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day [/fig_ref] , and five at the training center (nos. 4−8, [fig_ref] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day [/fig_ref]. No discomfort or work interference was reported from the samplers regardless of configuration. All extracted samplers contained measurable levels of PAHs, 12 of which are on the EPA priority list.In addition, two OPAHs (benzofluorenone and fluorenone) were detected and quantifi-able in both occupational settings. OPAHs are not typically monitored in asphalt exposures, so this represents some of the first evidence of a potential data gap in occupational exposure. Total PAHs ranged from 230 to 4600 ng/PSD [fig_ref] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day [/fig_ref] and trip blank PAH concentrations were all below 11 ng/PSD. Therefore, wristbands were extremely sensitive even after only 8 h of exposure, and individual PAH concentrations fromsilicone PSDs exceeded instrument detection limits from 2 to over 1400 fold. In addition, blanks had PAHs below However, levels of these two PAHs in blanks were negligible, considering average background from either PAH was nominally 3-fold lower or more than any individual's deployed sample. There were no detectable OPAHs in any blank. Individual PAH surrogate recoveries ranged from 53 to 122% (average = 91%, median = 94%) while OPAH surrogate recoveries ranged 64−120% (average = 83%; median = 82%). While the chromatography was easier to interpret for lapel and stacked designs, all PAHs and OPAHs were able to be identified and accurately quantified in all three configurations at all exposure durations. Although sample sizes are small at either site, there is no statistical difference between configurations (p > 0.05, power <0.8, SI . However, in some cases both single and stacked designs had lower concentrations than lapels for some roofers (participants nos. 5, and 7, [fig_ref] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day [/fig_ref] , and after reviewing participant questionnaires it was determined that these wristbands (either in stacked or single configurations) were worn underneath protective clothing. In the case of participant 8, who had a lower value for the lapel than other roofers, the survey data indicated that this lapel was covered as well. Not all discrepancies can be explained with protective clothing information (participant 4's stacked wristband was reportedly covered yet is the highest value for that individual), but in future occupational applications, it would be important to indicate how the sampler should be worn with respect to personal protection equipment. Even though PSDs are sequestering the vapor phase, it is likely that personal protection equipment impacts the level of exposures seen by restricting air flow with respect to a noncovered PSD. Further study would be needed to explore this idea, and this potential application of evaluating the effectiveness of protective clothing. Ultimately however, designs did not significantly differ, and all PSDs from each individual were pooled together to observe the trends described below.
Regarding temporal sensitivity, there was a significant difference between single day or multiday exposures (p < 0.05, power = 0.85, [fig_ref] Figure 4: Worksite PSDs [/fig_ref]. Interestingly, 22 out of 23 PAHs and OPAHs detected in the 40 h deployment were also detected in the 8 h deployment, further illustrating the capability of the sampler for typical 8 or 10 h time-weighted averages. Benzo[a]pyrene was not detected in the 8 h deployment, but benzo[e]pyrene was quantitated, refuting the inference that larger PAHs would not be able to be detected in a shorter exposure period. Additionally, because benzo[a]pyrene was just above reporting limits after nominally 40 h of exposure, it is likely that this compound was too low for our methods to quantitate at 8 h, rather than a failure of the sampler itself. Phenanthrene and alkylated phenanthrenes were the most common and most abundant individual PAHs [fig_ref] Figure 4: Worksite PSDs [/fig_ref]. Other atmospheric PAH profiles report the prominence of phenanthrene in hot asphalt exposure, [bib_ref] Road pavers' occupational exposure to asphalt containing waste plastic and tall oil..., Vaananen [/bib_ref] and it has the highest emission rate out of 14 PAHs measured in working asphalt. [bib_ref] Bitumen fumes: Review of work on the potential risk to workers and..., Binet [/bib_ref] Unexpectedly, naphthalene and alkylated homologues [fig_ref] Figure 4: Worksite PSDs [/fig_ref] were higher in 8-h over 40-h deployments. Differences in compound equilibrium between silicone and the atmosphere could explain naphthalene concentrations over time, and it is known that naphthalene is difficult to interpret with work-related exposure due to confounders such as cigarette smoking. [bib_ref] Investigation of PAH biomarkers in the urine of workers exposed to hot..., Sobus [/bib_ref] In fact, participants 2 and 3 did report cigarette use, while participant 1 did not. However, due to the small sample size, we are reluctant to over interpret the results here. At both occupational sites personal silicone samplers were worn for approximately one 8 h work day. While this small study cannot examine specific differences between worksites, we make the following casual observation about spatial sensitivity as detected by our passive sampler. Individuals at the rooftop site had similar profiles of PAH exposure, but differed in magnitude between participant 1 and participants 2−3 [fig_ref] Figure 4: Worksite PSDs [/fig_ref]. Survey information indicated participant 1 was a safety monitor at the worksite, while the other two participants were journeymen roofing professionals that reported directly handling hot asphalt. In another example, exposures were compared between occupational settings, and a significant difference was seen between study sites (p < 0.05, power = 0.99, [fig_ref] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day [/fig_ref]. The training center had a higher average PAH concentration than at the worksite (training center: 3040 ± 1090 ng/PSD; worksite: 800 ± 570 ng/PSD). Survey reports indicated that while hot asphalt was used in a similar manner at both sites, there was a difference between work enclosures. At the training center, hot asphalt is used to build a simulated roof at ground level in a semienclosed outdoor space. In contrast, hot asphalt was used on the rooftop only after the old roofing material was taken out, reducing some of the asphalt exposure. Taken together, spatial evidence supports the use of silicone wristbands as sensitive personal monitoring PSDs for exposures in a real world application. However, since we did not expect this level of sensitivity, additional work should be carried out to explore more specific differences between individual exposures.
Silicone personal samplers present an innovative sampling technology platform producing relevant, quantifiable data. By using these passive samplers, an atmospheric, time-weighted average concentration over an exposure period can be compared with exposure limits and compliance measurements through in situ calibration. Future work using isotope-labeled performance reference compounds to obtain in situ sampling rates will be done by infusing these compounds into PSDs prior to use. Studies utilizing this sampler are currently underway, and we hope this easy-to-wear and dynamic application of silicone may become a valuable tool to address challenges of the exposome and mixture toxicity.
## ■ associated content
## * s supporting information
Additional roofing information; extraction efficiency under two time treatments with four labeled PAHs [fig_ref] Figure 1: Examples of silicone personal sampling samplers [/fig_ref] ; photographic demonstration of postdeployment cleaning [fig_ref] Figure 2: Total ion chromatograms of wristband extracts through stages of cleaning and overspike... [/fig_ref] ; graphical representation of nonmetric multidimensional scaling of binary ambient wristband data [fig_ref] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day [/fig_ref] ; sun and shade 4 h exposures of infused silicone samplers [fig_ref] Figure 4: Worksite PSDs [/fig_ref] ; simulated transport conditions at three temperatures ; and sample configurations from eight hour training PSD exposure . This material is available free of charge via the Internet at http://pubs.acs.org.
## ■ author information
Corresponding Author *Phone: 541-737-8501; fax: 541-737-0497; e-mail: kim. [email protected].
[fig] Figure 1: Examples of silicone personal sampling samplers. (a) [/fig]
[fig] Figure 2: Total ion chromatograms of wristband extracts through stages of cleaning and overspike on the GC-MS. All chromatograms are scaled equally to easily show differences in chromatograms. (a) A wristband background with five rounds of ethyl acetate/methanol. (b) The addition of hexane to solvent precleaning drastically reduced total background inferences. Peaks here were identified as forms of siloxanes from mass spectral comparisons to NIST libraries. (c) Notable peaks of the overspike chromatogram are labeled with corresponding PAH abbreviations. [/fig]
[fig] Figure 3: Three silicone passive sampler designs sequestering PAHs in a single work day. (a) All samplers from both occupational settings are pictured including those from nos. 1−3 at a rooftop worksite, and nos. 4−8 at the training facility. The lapel corresponding to no. 3 was lost during the field deployment. Standard deviations here are derived from nondeployed wristbands (n = 5) representing laboratory and instrumental variability spiked with all target PAHs (average RSD: 2.30%). Blue stars represent silicone passive samplers that were reported as covered with protective clothing during exposure. (b) Overall exposures between sites differed significantly over an 8 h work period (p < 0.05). Standard deviations here are the result of all samplers pooled together from each participant. [/fig]
[fig] Figure 4: Worksite PSDs (all configurations) with sum (4a) and individual PAH exposure (4b−d) for a single (8 h) or multiday period (40 h). Individual profiles are scaled equally to observe differences in magnitude and profile between silicone samplers. Standard deviations are the result of PSDs pooled together from each participant. [/fig]
[table] Table 1: Compounds Identified from AMDIS Spectra against Chemical Libraries during Ambient Exposures a [/table]
|
The progress, prospects, and challenges of the use of non-coding RNA for diabetic wounds
Chronic diabetic wounds affect the quality of life of patients, resulting in significant social and economic burdens on both individuals and the health care system. Although treatment methods for chronic diabetic wounds have been explored, there remains a lack of effective treatment strategies; therefore, alternative strategies must be explored. Recently, the abnormal expression of non-coding RNA in diabetic wounds has received widespread attention since it is an important factor in the development of diabetic wounds. This article reviews the regulatory role of three common non-coding RNAs (microRNA [miRNA], long non-coding RNA [lncRNA], and circular RNA [circRNA]) in diabetic wounds and discusses the diagnosis, treatment potential, and challenges of non-coding RNA in diabetic wounds. This article provides insights into new strategies for diabetic wound diagnosis and treatment at the genetic and molecular levels.
# Introduction
Diabetes has become one of the main common diseases threatening public health. [bib_ref] IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for, Ogurtsova [/bib_ref] [bib_ref] Trends in cause-specific mortality among adults with and without diagnosed diabetes in..., Gregg [/bib_ref] Currently, there are 340 million individuals with diabetes worldwide, and this number is expected to increase to 440 million by 2030, resulting in 4.6 million deaths every year. [bib_ref] MicroRNA-296-5p promotes healing of diabetic wound by targeting sodium-glucose transporter 2 (SGLT2), Liu [/bib_ref] [bib_ref] Diabetes impairs wound healing by Dnmt1-dependent dysregulation of hematopoietic stem cells differentiation..., Yan [/bib_ref] Approximately 25% of all patients with diabetes suffer from diabetic foot ulcers (DFUs), one of the most common associated complications, which requires repeated surgical interventions and has become the main cause of non-traumatic amputation with a 5-year mortality rate of >50%; hence, it has caused a significant burden to patients. [bib_ref] Genetics of diabetes mellitus and diabetes complications, Cole [/bib_ref] [bib_ref] Dysfunctional Wound Healing in Diabetic Foot Ulcers: New Crossroads, Davis [/bib_ref] [bib_ref] Platelet-rich plasma plays an antibacterial, anti-inflammatory and cell proliferation-promoting role in an..., Li [/bib_ref] [bib_ref] Topical mevastatin promotes wound healing by inhibiting the transcription factor c-Myc via..., Sawaya [/bib_ref] [bib_ref] Circular RNA hsa_circ_0084443 Is Upregulated in Diabetic Foot Ulcer and Modulates Keratinocyte..., Wang [/bib_ref] [bib_ref] Diabetic foot disorders. A clinical practice guideline (2006 revision), Frykberg [/bib_ref] Imbalances in regulatory factor activity may lead to delayed healing and consequent formation of chronic wounds, [bib_ref] Wound repair and regeneration: mechanisms, signaling, and translation, Eming [/bib_ref] such as DFUs. At present, diabetic wounds are mainly treated by controlling excessive inflammation and promoting cell proliferation. Sadly, existing treatment strategies are not satisfactory, and few alternative options exist [fig_ref] Figure 1: Pathological characteristics of diabetic wounds [/fig_ref]. [bib_ref] Update on management of diabetic foot ulcers, Everett [/bib_ref] [bib_ref] Diabetic Foot Ulcers and Their Recurrence, Armstrong [/bib_ref] MicroRNAs (miRNAs) are endogenously expressed non-coding RNAs (ncRNAs) with a length of 18À22 nucleotides, and they contribute significantly to gene regulation at the post-transcriptional level. [bib_ref] Quantification of the differential expression levels of microRNA-203 in different degrees of..., Liu [/bib_ref] Studies have shown that miRNAs are involved in diabetic wound healing by regulating the functions of repair cells, such as im-mune cells, endothelial cells, fibroblasts, and keratinocytes. [bib_ref] Deduction of Novel Genes Potentially Involved in Keratinocytes of Type 2 Diabetes..., Lin [/bib_ref] Long ncRNAs (lncRNAs) are important to many cellular processes activated in wound healing, such as cell proliferation, migration, differentiation, and apoptosis, as well as angiogenesis and re-epithelialization. [bib_ref] Autologous blood transfusion augments impaired wound healing in diabetic mice by enhancing..., Guo [/bib_ref] Additionally, circular RNA (circRNA) helps regulate the physiological and pathological processes of many diseases, including diabetic wound healing. [bib_ref] Circular RNA hsa_circ_0084443 Is Upregulated in Diabetic Foot Ulcer and Modulates Keratinocyte..., Wang [/bib_ref] [bib_ref] High glucose-induced circHIPK3 downregulation mediates endothelial cell injury, Cao [/bib_ref] At present, there are few systematic and comprehensive reports describing the function of ncRNAs in diabetic wound healing. This review summarizes what is known about the regulatory role of three ncRNAs (miRNA, lncRNA, circRNA) in diabetic wound healing, as well as its application prospects and challenges.
ncRNA ncRNA refers to a type of RNA that does not encode protein. In the past, it has been shown to participate in a variety of biological processes and play important transcriptional and post-transcriptional regulatory functions. [bib_ref] Landscape of transcription in human cells, Djebali [/bib_ref] ncRNAs are divided into two categories based on the length of the RNA chain; lncRNA have lengths greater than 200 nucleotides, and small ncRNAs (snRNA) have lengths less than 200 nucleotides. snRNAs can be subdivided into miRNAs and circRNAs. [bib_ref] Noncoding RNAs in diabetes vascular complications, Beltrami [/bib_ref] miRNAs inhibit the expression of target genes by binding to the 3 0 untranslated region (3 0 UTR) of the target mRNA. [bib_ref] NFkappaB activation is essential for miR-21 induction by TGFb1 in high glucose..., Madhyastha [/bib_ref] The miRNA gene is processed and cleaved by RNA polymerase II (Pol II) enzyme and endonuclease Drosha III to produce pre-miRNA, which is transported by Exportin-5 protein to the cytoplasm, where it is digested by RNase III Dicer to produce mature miRNAs, which are loaded on Ago2 to form the large ribonucleoprotein effector complex, RNA-induced silencing complex (RISC).
In the diabetic wound microenvironment, hyperglycemia, inflammatory factors, and their downstream genes can bind to miRNA promoters or affect the methylation process, thereby controlling the miRNA levels, inhibiting the activity of Dicer enzyme, and reducing the expression of related miRNAs. [bib_ref] NFkappaB activation is essential for miR-21 induction by TGFb1 in high glucose..., Madhyastha [/bib_ref] [bib_ref] Hydrogen sulfide rescues high glucoseinduced migration dysfunction in HUVECs by upregulating miR-126-3p, Xue [/bib_ref] [bib_ref] Epigenetic Modification of MicroRNA-200b Contributes to Diabetic Vasculopathy, Singh [/bib_ref] [bib_ref] Inositol-Requiring Enzyme 1 Facilitates Diabetic Wound Healing Through Modulating MicroRNAs, Wang [/bib_ref] [bib_ref] Downregulation of miRNAs during Delayed Wound Healing in Diabetes: Role of Dicer, Bhattacharya [/bib_ref] Furthermore, the imbalanced miRNA can inhibit the expression of the target gene by interacting with the 3 0 UTR of the target mRNA, thereby inducing mRNA degra-dation and translation inhibition. [bib_ref] NFkappaB activation is essential for miR-21 induction by TGFb1 in high glucose..., Madhyastha [/bib_ref] [bib_ref] Origins and Mechanisms of miRNAs and siRNAs, Carthew [/bib_ref] [bib_ref] Overview of MicroRNA Biogenesis, Mechanisms of Actions, and Circulation, O'brien [/bib_ref] The interaction of miRNAs with other regions has also been reported, including 5 0 UTR, coding sequence, methylation modification, and gene promoters. [bib_ref] Diabetes impairs wound healing by Dnmt1-dependent dysregulation of hematopoietic stem cells differentiation..., Yan [/bib_ref] [bib_ref] Origins and Mechanisms of miRNAs and siRNAs, Carthew [/bib_ref] [bib_ref] Overview of MicroRNA Biogenesis, Mechanisms of Actions, and Circulation, O'brien [/bib_ref] lncRNAs lack an open reading frame and can regulate gene expression at multiple levels. [bib_ref] Long non-coding RNAs: insights into functions, Mercer [/bib_ref] lncRNAs play an epigenetic regulatory role by regulating methylation and histone modification; [bib_ref] Autologous blood transfusion augments impaired wound healing in diabetic mice by enhancing..., Guo [/bib_ref] [bib_ref] Diabetes Mellitus-Induced Long Noncoding RNA Dnm3os Regulates Macrophage Functions and Inflammation via..., Das [/bib_ref] they also interact with RNA molecules or protein complexes for transcriptional regulation [bib_ref] Topical mevastatin promotes wound healing by inhibiting the transcription factor c-Myc via..., Sawaya [/bib_ref] [bib_ref] Diabetes Mellitus-Induced Long Noncoding RNA Dnm3os Regulates Macrophage Functions and Inflammation via..., Das [/bib_ref] [bib_ref] Enhancer-associated long non-coding RNA LEENE regulates endothelial nitric oxide synthase and endothelial..., Miao [/bib_ref] and can act as a molecular sponge of miRNAs to regulate downstream target genes for post-transcriptional regulation. [bib_ref] The MSC-Derived Exosomal lncRNA H19 Promotes Wound Healing in Diabetic Foot Ulcers..., Li [/bib_ref] [bib_ref] Long noncoding RNA-antisense noncoding RNA in the INK4 locus accelerates wound healing..., He [/bib_ref] [bib_ref] Improved therapeutic effects on diabetic foot by human mesenchymal stem cells expressing..., Zhu [/bib_ref] In addition, lncRNAs can also affect protein expression and interfere with protein binding. [bib_ref] Screening and preliminary validation of T lymphocyte immunoregulation-associated long non-coding RNAs in..., Xu [/bib_ref] [bib_ref] Extracellular vesicle-mimetic nanovesicles transport LncRNA-H19 as competing endogenous RNA for the treatment..., Tao [/bib_ref] [bib_ref] lncRNA MALAT1 Accelerates Wound Healing of Diabetic Mice Transfused with Modified Autologous..., Liu [/bib_ref] [bib_ref] Long non-coding RNA Lethe regulates hyperglycemia-induced reactive oxygen species production in macrophages, Zgheib [/bib_ref] [bib_ref] WAKMAR2, a Long Noncoding RNA Downregulated in Human Chronic Wounds, Modulates Keratinocyte..., Herter [/bib_ref] Recent studies have shown Under the action of various factors, the function of wound-repair cells is impaired, and the extracellular matrix and its cytokines are degraded. This leads to chronic inflammation of the wound, vascular dysfunction, and obstruction of re-epithelialization and delayed extracellular matrix remodeling. All of these events also promote and reinforce each other to eventually inhibit diabetic wound repair. that lncRNAs are key regulators of cell proliferation, differentiation, apoptosis, and other cell biological processes, thereby regulating wound healing [fig_ref] Figure 3: Overview of the mechanisms of non-coding RNA [/fig_ref]. [bib_ref] Noncoding RNAs in diabetes vascular complications, Beltrami [/bib_ref] [bib_ref] Long non-coding RNAs: insights into functions, Mercer [/bib_ref] [bib_ref] The MSC-Derived Exosomal lncRNA H19 Promotes Wound Healing in Diabetic Foot Ulcers..., Li [/bib_ref] [bib_ref] Mechanisms of long noncoding RNA function in development and disease, Schmitz [/bib_ref] In circRNA, the 5 0 and 3 0 ends are connected to form a complete closed-loop structure. Therefore, circRNAs are resistant to degradation by the RNA exonuclease, making them more stable than linear RNAs. [bib_ref] Exosomes derived from mmu_circ_0000250-modified adipose-derived mesenchymal stem cells promote wound healing in..., Shi [/bib_ref] [bib_ref] Circular RNA: An emerging non-coding RNA as a regulator and biomarker in..., Chen [/bib_ref] circRNAs can regulate gene transcription mediated by STAT3 nuclear translocation and act as a molecular sponge of miRNA to mitigate its inhibitory effect on the downstream target genes [bib_ref] High glucose-induced circHIPK3 downregulation mediates endothelial cell injury, Cao [/bib_ref] [bib_ref] Exosomes derived from mmu_circ_0000250-modified adipose-derived mesenchymal stem cells promote wound healing in..., Shi [/bib_ref] [bib_ref] The Circular RNA Interacts with STAT3, Increasing Its Nuclear Translocation and Wound..., Yang [/bib_ref]. They can also exhibit a gene regulatory function by competing with linear RNA splicing and a protein coding ability. [bib_ref] Circular RNA: An emerging non-coding RNA as a regulator and biomarker in..., Chen [/bib_ref] [bib_ref] Guidance of circular RNAs to proteins' behavior as binding partners, Luo [/bib_ref] Furthermore, they can also interact with RNA-binding proteins, [bib_ref] Circ_0075932 in adipocytederived exosomes induces inflammation and apoptosis in human dermal keratinocytes..., Zhang [/bib_ref] RNA-processing proteins, [bib_ref] The Circular RNA Interacts with STAT3, Increasing Its Nuclear Translocation and Wound..., Yang [/bib_ref] [bib_ref] Circ_0075932 in adipocytederived exosomes induces inflammation and apoptosis in human dermal keratinocytes..., Zhang [/bib_ref] transcription factors, and proteases to regulate their localization and stability. [bib_ref] Guidance of circular RNAs to proteins' behavior as binding partners, Luo [/bib_ref] [bib_ref] Circular intronic long noncoding RNAs, Zhang [/bib_ref] [bib_ref] Foxo3 circular RNA retards cell cycle progression via forming ternary complexes with..., Du [/bib_ref] Recent studies have shown that circRNA dysregulation may facilitate the occurrence and development of diabetes [fig_ref] Figure 4: Mechanisms of circRNA function circRNAs can regulate gene transcription mediated by STAT3... [/fig_ref]. [bib_ref] High glucose-induced circHIPK3 downregulation mediates endothelial cell injury, Cao [/bib_ref] [bib_ref] Circ_0075932 in adipocytederived exosomes induces inflammation and apoptosis in human dermal keratinocytes..., Zhang [/bib_ref] [bib_ref] Profiling and functional analysis of differentially expressed circular RNAs in high glucose-induced..., Jin [/bib_ref] miRNAs, lncRNAs, and circRNAs can be employed as regulatory factors in the healing of diabetic wounds. miRNAs tend to act directly on mRNA of target genes and negatively regulate them, whereas lncRNAs can bind to downstream target gene mRNA and can also be used as competing endogenous RNA (ceRNA) to competitively bind miRNAs to achieve transcriptional-or post-transcriptional-level regulation. circRNA is mainly used as a molecular sponge of miRNAs to competitively inhibit miRNAs. lncRNA/circRNA/miRNA cross-link to form a huge and complex regulatory network that alters gene expression, thereby regulating multiple stages of diabetic wound repair.
The regulatory role of ncRNA in diabetic wound healing Diabetic wounds are a type of chronic wounds associated with hyperglycemia. Their pathological changes primarily include chronic inflammation, vascular dysfunction, delayed re-epithelialization, and disorder of the extracellular matrix (ECM) remodeling. They are attributed to excessive accumulation of advanced glycation end . Biogenesis and regulation mechanism of miRNA function (1) The miRNA gene was cleaved and processed by RNA polymerase II enzyme and endonuclease Drosha III to form pre-miRNA. The Exportin-5 protein then transports the pre-miRNA to the cytoplasm where it is digested by Dicer to produce mature miRNA, which is loaded on Ago2 to form RISC. Then, it inhibits transcription (A), regulates the degradation of mRNA (B), combine with 5 0 UTR (C) and coding sequence (D) ,etc. (2) During hyperglycemia and inflammation, the effects of p65 and miR-21 promoter are enhanced, whereas miR-21 expression is induced. RNase IRE1alpha reduces the levels of miR-200 and miR-466 to that of pre-miRNA. In addition, hyperglycemia can change the expression of miRNA epigenetics by affecting the DNMT methylation process. The expression-imbalanced miRNA can inhibit the expression of the target gene by interacting with the 3 0 untranslated region (3 0 UTR) of the target mRNA, inducing mRNA degradation, and inhibiting translation. In addition, miRNAs can also interact with 5 0 UTR and gene promoters and induce methylation modification. products (AGEs), 47-50 excessive oxidative stress (OS), 4,50-52 persistent inflammatory response, and imbalance of signal pathway activation. [bib_ref] Diabetes impairs wound healing by Dnmt1-dependent dysregulation of hematopoietic stem cells differentiation..., Yan [/bib_ref] [bib_ref] MicroRNA-155 Inhibition Promoted Wound Healing in Diabetic Rats, Ye [/bib_ref] [bib_ref] miR-5591-5p regulates the effect of ADSCs in repairing diabetic wound via targeting..., Li [/bib_ref] [bib_ref] MicroRNA miR-27b rescues bone marrow-derived angiogenic cell function and accelerates wound healing..., Wang [/bib_ref] [bib_ref] Long non-coding RNA MALAT1 regulates hyperglycaemia induced inflammatory process in the endothelial..., Puthanveetil [/bib_ref] Hence, they hinder immune cell function, 4,47,53 induce excessive inflammation, and damage repair cells around the wound, [bib_ref] miRNA-221-3p in Endothelial Progenitor Cell-Derived Exosomes Accelerates Skin Wound Healing in Diabetic..., Xu [/bib_ref] [bib_ref] Inhibition of microRNA-34a mediates protection of thymosin beta 4 in endothelial progenitor..., Chen [/bib_ref] [bib_ref] MiR-155 targets PTCH1 to mediate endothelial progenitor cell dysfunction caused by high..., Gao [/bib_ref] [bib_ref] LPS-preconditioned mesenchymal stromal cells modify macrophage polarization for resolution of chronic inflammation..., Ti [/bib_ref] all of which are accompanied with significant changes in ECM and other components, 56-59 thereby destroying the microenvironment of wound healing. Under the combined action of several complex factors, diabetic wounds can induce and exacerbate other diabetic wounds, thereby hindering the healing process of skin defects in patients with diabetes.
Previous studies have confirmed that, compared with normal wounds, the microenvironment of diabetic wounds can affect the expression level of ncRNA, which can reverse the diabetic wound microenvironment. The pathophysiological changes associated with diabetic wounds may be induced by the expression of ncRNA and its regulation imbalance of the downstream genes, which is caused by hyperglycemia, OS, inflammatory factors, and AGEs, thereby impairing cell function, resulting in chronic inflammation of the wound; angiogenesis blocked, re-epithelialization delays; and disorders of ECM remodeling. [bib_ref] miR-5591-5p regulates the effect of ADSCs in repairing diabetic wound via targeting..., Li [/bib_ref] [bib_ref] Inhibition of microRNA-34a mediates protection of thymosin beta 4 in endothelial progenitor..., Chen [/bib_ref] [bib_ref] Long non-coding RNA MALAT1 regulates hyperglycaemia induced inflammatory process in the endothelial..., Puthanveetil [/bib_ref] [bib_ref] Identification of Specific miRNAs in Neutrophils of Type 2 Diabetic Mice: Overexpression..., Umehara [/bib_ref] [bib_ref] MicroRNA-129 and -335 Promote Diabetic Wound Healing by Inhibiting Sp1-Mediated MMP-9 Expression, Wang [/bib_ref] [bib_ref] Downregulation of endothelial microRNA-200b supports cutaneous wound angiogenesis by desilencing GATA binding..., Chan [/bib_ref] [bib_ref] Mesenchymal stem cells correct impaired diabetic wound healing by decreasing ECM proteolysis, Xu [/bib_ref] On the contrary, when reversing its expression, NcRNA can induce the expression of related downstream genes, which significantly reduce both the damage of the diabetic wound microenvironment to cells and the level of OS and reactive oxygen species (ROS). This ultimately will inhibit cell apoptosis, improve the level of inflammation, and promote the proliferation, migration, and differentiation of endothelial cells, keratinocytes, and fibroblasts to accelerate the process of angiogenesis, re-epithelialization, and ECM remodeling, which in turn accelerate wound healing [fig_ref] Figure 5: Regulatory effects of certain common ncRNAs on the pathology of diabetic wounds... [/fig_ref]. [bib_ref] lncRNA MALAT1 Accelerates Wound Healing of Diabetic Mice Transfused with Modified Autologous..., Liu [/bib_ref] [bib_ref] miR-5591-5p regulates the effect of ADSCs in repairing diabetic wound via targeting..., Li [/bib_ref] [bib_ref] MicroRNA miR-27b rescues bone marrow-derived angiogenic cell function and accelerates wound healing..., Wang [/bib_ref] [bib_ref] Long non-coding RNA MALAT1 regulates hyperglycaemia induced inflammatory process in the endothelial..., Puthanveetil [/bib_ref] [bib_ref] Engineered Human Adipose Stem-Cell-Derived Exosomes Loaded with miR-21-5p to Promote Diabetic Cutaneous..., Lv [/bib_ref] [bib_ref] Shedding light on miR-26a: Another key regulator of angiogenesis in diabetic wound..., Zgheib [/bib_ref] [bib_ref] MicroRNA-21-3p accelerates diabetic wound healing in mice by downregulating SPRY1, Wu [/bib_ref] [bib_ref] miR-27-3p inhibition restore fibroblasts viability in diabetic wound by targeting NOVA1, Zhang [/bib_ref] miRNAs The microenvironment of diabetic wounds can induce an imbalance in miRNA expression by affecting the promoter, RNase activity, and methylation process. [bib_ref] NFkappaB activation is essential for miR-21 induction by TGFb1 in high glucose..., Madhyastha [/bib_ref] [bib_ref] Epigenetic Modification of MicroRNA-200b Contributes to Diabetic Vasculopathy, Singh [/bib_ref] [bib_ref] Inositol-Requiring Enzyme 1 Facilitates Diabetic Wound Healing Through Modulating MicroRNAs, Wang [/bib_ref] [bib_ref] Downregulation of miRNAs during Delayed Wound Healing in Diabetes: Role of Dicer, Bhattacharya [/bib_ref] Moreover, the abnormally expressed miRNA is involved in various processes of diabetic wound healing, which significantly affects the process of normal wound repair. [bib_ref] Downregulation of miRNAs during Delayed Wound Healing in Diabetes: Role of Dicer, Bhattacharya [/bib_ref] [bib_ref] Maggot excretions/secretions promote diabetic wound angiogenesis via miR18a/19a -TSP-1 axis, Wang [/bib_ref] [bib_ref] Integrative analysis of miRNA and mRNA paired expression profiling of primary fibroblast..., Liang [/bib_ref]
## Regulation of wound inflammation
The transition from the inflammation phase to the proliferative phase during the wound-healing process is critical. [bib_ref] Wound healing and its impairment in the diabetic foot, Falanga [/bib_ref] [bib_ref] Exosome-Guided Phenotypic Switch of M1 to M2 Macrophages for Cutaneous Wound Healing, Kim [/bib_ref] Previous studies have shown that the imbalance of miRNA expression may explain the imbalance of wound inflammation [fig_ref] Table 1: Functions and mechanisms of miRNAs in diabetic wounds [/fig_ref].
## Review
## Anti-inflammatory effects
Neutrophils initiate the first stage of acute inflammation. The intrinsic defect of diabetic neutrophils may be one of the causes of chronic inflammation of DFUs, whereas microRNA (miR)-129-2-3p may be involved in neutrophil dysfunction regulation. In a previous study, it was reported that although the number of neutrophils in type 2 diabetic (T2D) mice was significantly higher than that in nondiabetic mice, the expression of miR-129-2-3p was significantly lower in the T2D mice, resulting in delayed wound healing. Contrarily, overexpression of miR-129-2-3p in the wounds of T2D mice promoted wound healing. It is possible that miR-129-2-3 directly inhibits the expression of Casp6 and Ccr2, which are involved in apoptosis, inflammatory response, and chemotaxis, as well as the regulation of phagocytosis and endocytosis. [bib_ref] Identification of Specific miRNAs in Neutrophils of Type 2 Diabetic Mice: Overexpression..., Umehara [/bib_ref] Macrophage infiltration can exert an anti-inflammatory effect and remove cell debris to prevent infection in the early stages of wound healing. The late inflammatory stage transition of macrophages between pro-inflammatory and anti-inflammatory phenotypes is key to entering to the proliferative phase. [bib_ref] Exosome-Guided Phenotypic Switch of M1 to M2 Macrophages for Cutaneous Wound Healing, Kim [/bib_ref] [bib_ref] Wound repair and regeneration, Reinke [/bib_ref] miRNAs play an anti-inflammatory regulatory role mainly by regulating the number, phenotype, and differentiation of macrophages. For example, miR-let-7b regulates macrophage polarization by inhibiting the Toll-like receptor 4 (TLR4)/nuclear factor kB (NF-kB)/STAT3/a serine/threonine kinase (AKT) signal axis. [bib_ref] LPS-preconditioned mesenchymal stromal cells modify macrophage polarization for resolution of chronic inflammation..., Ti [/bib_ref] Similarly, Nox-2 increased OS, which resulted in a decrease in the expression of miR-let-7d-3p in hematopoietic stem cells of T2D mice. miR-let-7d-3p directly upregulated DNA methyltransferase (Dnmt)1, a key enzyme that mediates DNA methylation, causing Notch1, PU.1, and Klf4 to be highly methylated, which resulted in the reduction of the number of macrophages in the wound and facilitated the polarization of macrophages to M1, thus aggravating a severe reaction. miR-let-7d-3p epigenetically inhibited the differentiation of hematopoietic stem cells into monocytes/macrophages and regulated the balance of the M1/M2 ratio in wounds, thereby affecting wound healing. [bib_ref] Diabetes impairs wound healing by Dnmt1-dependent dysregulation of hematopoietic stem cells differentiation..., Yan [/bib_ref] Interestingly, another study reported that fibroblasts in wound-granulation tissue are derived from myeloid cells such as macrophages. Keratinocyte-derived miR-21 is packaged in extracellular vesicles (EVs) and positively regulates the plasticity of wound macrophages, reverses the barriers to the conversion of myeloid cells to fibroblast-like cells in diabetic wounds, and reduces wound infiltration by macrophages, thus exerting an active anti-inflammatory effect. [bib_ref] Direct conversion of injury-site myeloid cells to fibroblast-like cells of granulation tissue, Sinha [/bib_ref] Keratinocytes are an important epidermal cell. Previous studies have shown that the hyperglycemia environment in diabetic skin can enhance OS and increase the inflammatory response of keratinocytes, thus stagnating the wound-healing process in the inflammatory phase. [bib_ref] Platelet-rich plasma plays an antibacterial, anti-inflammatory and cell proliferation-promoting role in an..., Li [/bib_ref] [bib_ref] MicroRNA-132 with Therapeutic Potential in Chronic Wounds, Li [/bib_ref] Studies have confirmed that miR-132 is the key miRNA that determines the transition from inflammation to proliferation in wounds. Both in vivo and in vitro experiments have confirmed that miR-132 knockout causes severe inflammation, inhibits keratinocyte proliferation, and delays wound healing. [bib_ref] MicroRNA-132 enhances transition from inflammation to proliferation during wound healing, Li [/bib_ref] On the contrary, the topical use of miR-132 mimics on diabetic wounds can effectively improve the inflammatory response and accelerate wound closure.
Through transcriptome analysis, it was shown that miR-132 plays an anti-inflammatory role in promoting diabetic wound healing by regulating NF-kB, nucleotide-binding oligomerization domain (NOD)-like receptors, TLRs, tumor necrosis factor (TNF) signaling pathways, and other inflammation-related signaling pathways [bib_ref] MicroRNA-132 with Therapeutic Potential in Chronic Wounds, Li [/bib_ref]. In addition, Li et al. [bib_ref] Platelet-rich plasma plays an antibacterial, anti-inflammatory and cell proliferation-promoting role in an..., Li [/bib_ref] reported that miR-21 has an anti-inflammatory effect by targeting PDCD4 to regulate the NF-kB signaling pathway. In other words, miRNAs convey anti-inflammatory effects by regulating the inflammatory signals of keratinocytes and reducing the production of inflammatory factors.
Finally, inflammatory factors and chemokines also play an indispensable role in the healing of diabetic wounds. Previous studies have shown that miR-124a and miR-125b inhibit the accumulation of chemokines and cytokines in diabetic wounds, exert anti-inflammatory effects, and accelerate wound healing. [bib_ref] Human fibrocyte-derived exosomes accelerate wound healing in genetically diabetic mice, Geiger [/bib_ref] In vivo and in vitro studies have shown that miR-497 can reduce the levels interleukin (IL)-1b, IL-6, TNF-a, and other pro-inflammatory factors and promote wound epithelialization and granulation tissue formation, thereby accelerating the diabetic wounds-healing process. [bib_ref] Accelerated wound healing in diabetic mice by miRNA-497 and its anti-inflammatory activity, Ban [/bib_ref] Furthermore, miR-146a was significantly downregulated during the healing process of diabetic wounds, leading to a significant increase in the downstream pro-inflammatory targets IRAK1, TRAF6, and NF-kB and an increase in IL-6 and MIP-2 gene expression. In addition, mesenchymal stem cells (MSCs) accelerate wound healing and relieve inflammation by increasing the expression of miR-146a in diabetic wounds. [bib_ref] The role of microRNA-146a in the pathogenesis of the diabetic wound-healing impairment:..., Xu [/bib_ref] [bib_ref] Use of Cerium Oxide Nanoparticles Conjugated with MicroRNA-146a to Correct the Diabetic..., Zgheib [/bib_ref] miR-146a can be coupled with cerium oxide nanoparticles to reduce wound inflammation, promote angiogenesis, and accelerate wound healing. [bib_ref] Use of Cerium Oxide Nanoparticles Conjugated with MicroRNA-146a to Correct the Diabetic..., Zgheib [/bib_ref] In conclusion, miRNAs can directly reduce the production of inflammatory factors and chemokines to exert anti-inflammatory effects.
## Pro-inflammatory effects
Overexpression of miRNAs may be one reason for the chronic inflammation of diabetic wounds. miR-155, a pro-inflammatory miRNA mainly expressed in immune cells, is significantly upregulated in diabetic skin, and the use of miR-155 inhibitors on wounds can reduce wound inflammation. miR-155 inhibition reduced the infiltration of inflammatory cells such as neutrophils, macrophages, and T cells on the wound, whereas increasing the number of M2 macrophages. [bib_ref] MicroRNA-155 Inhibition Promoted Wound Healing in Diabetic Rats, Ye [/bib_ref] The inhibitory effect on its target gene (FGF-7) was reversed, stimulating keratinocyte migration and proliferation, increasing vascular remodeling, and accelerating wound closure. [bib_ref] microRNA-155 inhibition restores Fibroblast Growth Factor 7 expression in diabetic skin and..., Moura [/bib_ref] In addition, hyperglycemia can induce the expression of NOX2 and the production of ROS by activating the miR-21/phosphatidylinositol 3-kinase (PI3K)/NOX2/ROS signaling pathway, thus increasing the inflammatory response. These results suggest that an imbalance in miR-21 expression may be one of the reasons for the continuous polarization and abnormal inflammation of M1 macrophages in diabetic wounds. [bib_ref] Role of microRNA-21 and Its Underlying Mechanisms in Inflammatory Responses in Diabetic..., Liechty [/bib_ref] Previous studies have shown that miR-15b-5p can deter DNA repair and pro-inflammatory effects by inhibiting downstream target genes, such as DNA repair genes WEE1, RAD50, MSH2, and KIT, and trauma inflammatory response target genes IKBKB and FGF2. [bib_ref] Staphylococcus aureus Triggers Induction of miR-15B-5P to Diminish DNA Repair and Deregulate..., Ramirez [/bib_ref] Therefore, imbalanced miRNAs play a pro-inflammatory role in wound repair by regulating the function of immune cells and activating inflammatory response signals.
## Angiogenesis regulation
Angiogenesis is another key process in wound repair, which is necessary for the delivery and supply of nutrients to the wound so as to promote the healing process. [bib_ref] Acceleration of Diabetic Wound Healing with PHD2-and miR-210-Targeting Oligonucleotides, Dallas [/bib_ref] [bib_ref] Angiogenesis and wound repair: when enough is enough, Dipietro [/bib_ref] miRNA can participate in the process of angiogenesis by regulating the function of endothelial cells and the expression of angiogenesis-related factors. miRNA may promote anti-angiogenesis or angiogenesis in wound healing.
## Angiogenesis promotion
miRNAs promote vascular endothelial cell proliferation, migration, and angiogenesis by regulating angiogenic growth factors. In contrast, the downregulation of the expression of some miRNAs in diabetic wounds may inhibit angiogenesis. TSP-1 has been shown to inhibit the adhesion and proliferation of several types of endothelial cells as well as tubular formation. [bib_ref] Maggot excretions/secretions promote diabetic wound angiogenesis via miR18a/19a -TSP-1 axis, Wang [/bib_ref] [bib_ref] Inhibition of tumor growth by systemic treatment with thrombospondin-1 peptide mimetics, Reiher [/bib_ref] In damaged bone marrow-derived angiogenic cells (BMACs) in a diabetic mouse model (db/db), miR-27b expression was decreased, whereas TSP-1 expression was increased. miR-27b improved the proliferation, adhesion, and angiogenesis of db/db BMACs, reduced OS, and protected BMAC function. Additionally, miR-27b directly targeted TSP-1, TSP-2, p66shc, and semaphorin6 to promote angiogenesis and increase blood perfusion and accelerate the closure of diabetic skin wounds. [bib_ref] MicroRNA miR-27b rescues bone marrow-derived angiogenic cell function and accelerates wound healing..., Wang [/bib_ref] Similarly, miR-18a/19a can accelerate the angiogenesis of diabetic wounds by downregulating TSP-1 and promoting endothelial cell migration and tubular formation. [bib_ref] Maggot excretions/secretions promote diabetic wound angiogenesis via miR18a/19a -TSP-1 axis, Wang [/bib_ref] Similarly, miR-23a inhibited the expression of IRF-1 and reduced the expression of inducible nitric oxide synthase (iNOS) and the angiogenesis-related factor vascular endothelial growth factor (VEGF), thus inhibiting angiogenesis and the healing process of diabetic wounds.Another study showed that miR-221-3p increased the expression of VEGF and promoted cell proliferation and angiogenesis. Bioinformatics analysis indicated that miR-221-3p may be involved in AGE-receptor for AGE (RAGE) signaling, cell cycle, and p53 signaling in diabetes. [bib_ref] miRNA-221-3p in Endothelial Progenitor Cell-Derived Exosomes Accelerates Skin Wound Healing in Diabetic..., Xu [/bib_ref] miR-126 is an important molecule that regulates angiogenesis. It has been shown that the expression of miR-126 is significantly reduced in the peripheral blood of DFU patients. The abnormal expression of . circRNAs can also regulate genes by competing with linear RNA splicing (E), while having protein-coding roles (D). In addition, they can also bind to certain proteins, such as circ_0075932 and hsa_circ_0084443, and interact with others to play a regulatory role (C).
miR-126 is closely related to the pathological changes in DFUs. [bib_ref] miR-126 as a Therapeutic Agent for Diabetes Mellitus, Pishavar [/bib_ref] Overexpression of miR-126 can reduce the production of ROS and apoptosis in the blood-brain barrier of patients with diabetes, activate the migration and recruitment of endothelial cells to the injured site, and accelerate tissue repair. [bib_ref] Increasing the miR-126 expression in the peripheral blood of patients with diabetic..., Zhang [/bib_ref] Under high glucose (HG) levels, miR-126-3p in human umbilical vein endothelial cells (HUVECs) was downregulated, which increased the expression of DNMT1, an anti-angiogenic DNMT, leading to HUVEC migration dysfunction. [bib_ref] Hydrogen sulfide rescues high glucoseinduced migration dysfunction in HUVECs by upregulating miR-126-3p, Xue [/bib_ref] Previous studies have shown that exosomes (exos) derived from synovium MSCs overexpressing miR-126-3p can significantly activate PI3K/AKT and mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) pathways, which can activate endothelial cell proliferation, migration, and angiogenesis in vitro and stimulate angiogenesis and re-epithelialization in vivo to accelerate wound healing. [bib_ref] Chitosan Wound Dressings Incorporating Exosomes Derived from MicroRNA-126-Overexpressing Synovium Mesenchymal Stem Cells..., Tao [/bib_ref] Finally, miRNAs can also promote angiogenesis by regulating stem cell functions. Previous studies have shown that human adiposederived stem cells (hADSCs) are damaged in patients with diabetes, which is manifested by a decrease in the expression of miR-1248 and an increase in the expression of CITED2, an inhibitor of hypoxia-inducible factor 1a (HIF-1a). CITED2 also regulates stem cell activity and differentiation and inhibits angiogenesis, thus affecting the secretion of growth factors and cell proliferation and delaying wound healing. [bib_ref] Diabetes-induced glucolipotoxicity impairs wound healing ability of adipose-derived stem cells-through the miR-1248/CITED2/HIF-1a..., Xiao [/bib_ref] Therefore, miRNAs mainly modulate the function of endothelial cells, change the expression of angiogenic factors, and act on regulatory signals to promote angiogenesis.
## Anti-angiogenesis
Abnormal expression of miRNAs may be conducive to the inhibition of angiogenesis in diabetic wounds. The expression of some antiangiogenic miRNAs is significantly increased in diabetic wounds; contrarily, the inhibition of these miRNAs may significantly accelerate wound healing. Studies have found that inhibiting the expression of anti-angiogenic miR-92a may improve blood perfusion in ischemic diseases. MRG-110, an miR-92a inhibitor, can be administered intracutaneously in db/db mouse wounds to accelerate the formation and re-epithelialization of granulation tissue without causing systemic toxic damage. [bib_ref] A synthetic microRNA-92a inhibitor (MRG-110) accelerates angiogenesis and wound healing in diabetic..., Gallant-Behm [/bib_ref] A similar effect was observed in the wound model of piglets. [bib_ref] A synthetic microRNA-92a inhibitor (MRG-110) accelerates angiogenesis and wound healing in diabetic..., Gallant-Behm [/bib_ref] Lucas et al. [bib_ref] Light-inducible antimiR-92a as a therapeutic strategy to promote skin repair in healing-impaired..., Lucas [/bib_ref] developed a light-induced anti-miR-92a therapeutic approach to promote healing in diabetic mice using light-resistant protective groups. The result of their study showed that the target of light-induced anti-miR-92a was more accurate and inhibited the expression of miR-92a in skin but did not significantly affect the expression of miR-92a in other organs. Light activation of anti-miR-92a stimulated wound cell proliferation and angiogenesis and significantly accelerated the healing of diabetic mice, and its effect was similar to that of traditional anti-miRNAs. [bib_ref] Light-inducible antimiR-92a as a therapeutic strategy to promote skin repair in healing-impaired..., Lucas [/bib_ref] In addition, miR-26a can inhibit the proliferation and migration of endothelial cells and inhibit angiogenesis by downregulating the target gene SMAD1 and upregulating the cell cycle inhibitor p27. In contrast, local administration of the miR-26a inhibitor locked nucleic acid (LNA)-anti-miR-26a to wounds can induce angiogenesis and accelerate wound healing, but these effects are unrelated to changes in the ratio of M1/M2 macrophages. [bib_ref] Regulation of impaired angiogenesis in diabetic dermal wound healing by microRNA-26a, Icli [/bib_ref] Similarly, the inhibition of miR-152-3p can upregulate phosphatase and tensin (PTEN) and promote the healing of diabetic wounds. [bib_ref] Inhibition of miRNA-152-3p enhances diabetic wound repair via upregulation of PTEN, Xu [/bib_ref] Interestingly, the expression of circulating exos or miR-20b-5p in T2D patients was significantly upregulated, indicating that miR-20b-5p or patients with diabetes exos can hinder angiogenesis and delay wound healing; contrarily, knocking out miR-20b-5p accelerated wound healing. miR-20b-5p may inhibit the angiogenesis of HUVECs by inhibiting Wnt9b/b-catenin signaling. [bib_ref] Circulating Exosomal miR-20b-5p Inhibition Restores Wnt9b Signaling and Reverses Diabetes-Associated Impaired Wound..., Xiong [/bib_ref] miRNAs can exert an anti-vascular effect by regulating the expression of angiogenic factors. SDF-1a is commonly targeted by miRNAs of the miR-23 family. Studies have shown that SDF-1a was significantly decreased in the blood and tissue sections of DFU patients, and the expression of miR-23a and miR-23b was downregulated, whereas the expression of miR-23c was increased. Thus, elevated miR-23c negatively regulated SDF-1a and inhibited angiogenesis. miR-23c is also involved in the downregulation of the expression of other angiogenic factors, such as endothelial NOS (eNOS), HIF-1a, and VEGF, thus inhibiting angiogenesis. [bib_ref] miR-23c regulates wound healing by targeting stromal cell-derived factor-1a (SDF-1a/CXCL12) among patients..., Amin [/bib_ref] miR-205-5p inhibits VEGF protein translation by interacting with the 3 0 UTR of VEGF mRNA. In contrast, anti-miR-205-5p significantly increased the expression of VEGF in MSCs, thus promoting angiogenesis. [bib_ref] Suppression of microRNA-205-5p in human mesenchymal stem cells improves their therapeutic potential..., Zhu [/bib_ref] Previous studies have shown that the expression of angiogenic factors HIF-1a and VEGF can be downregulated by miR-217, thus hindering the healing of foot ulcers in mice; 88 miR-217 is upregulated in the serum of DFU patients and positively correlated with Wagner grade. Further research showed that HIF-1a is the direct target gene of miR-217. Inhibition of miR-217 can activate the HIF-1a/VEGF pathway, reduce inflammation, and promote angiogenesis. [bib_ref] Expression of miR-217 and HIF-1a/VEGF pathway in patients with diabetic foot ulcer..., Lin [/bib_ref] miR-15b is considered a key negative regulator of angiogenesis and mainly inhibits the expression of HIF-1a. [bib_ref] The role of microRNA-15b in the impaired angiogenesis in diabetic wounds, Xu [/bib_ref] [bib_ref] Effects of the antagomiRs 15b and 200b on the altered healing pattern..., Pizzino [/bib_ref] In addition, it was reported that miR-615-5p may inhibit the VEGF/AKT/eNOS signaling pathway and inhibit angiogenesis by targeting insulin growth factor 2 (IGF2) and RASSF2 in endothelial cells. Meanwhile, animal experiments found that the miR-615-5 inhibitor significantly increased angiogenesis, granulation tissue thickness, and woundclosure rate in db/db mice. [bib_ref] MicroRNA-615-5p Regulates Angiogenesis and Tissue Repair by Targeting AKT/eNOS (Protein Kinase B/Endothelial..., Icli [/bib_ref] These miRNAs inhibit angiogenesis in tissue repair by inhibiting VEGF-related signaling pathways.
Previous studies have found that miR-200b was downregulated immediately and temporarily in endothelial cells around the wound, thus eliminating the negative regulation of miR-200b on GATA2 and VEGFR2 and initiating the process of angiogenesis; however, in diabetic wounds, the TNF-a-mediated miR-200b-GATA2-VEGFR2 signal axis exerted an anti-angiogenic effect and delayed wound healing, which was reversed after neutralizing TNF-a. [bib_ref] Downregulation of endothelial microRNA-200b supports cutaneous wound angiogenesis by desilencing GATA binding..., Chan [/bib_ref] miR-200b can also initiate diabetic vascular disease through epigenetic modification; that is, HG reduces DNMT1 and DNMT3A, key enzymes for epigenetic modification, leading to endothelial dysfunction, which encompasses decreased endothelial nitric oxide, decreased low-density lipoprotein (LDL) uptake, and Matrigel tube-formation disorder. [bib_ref] Epigenetic Modification of MicroRNA-200b Contributes to Diabetic Vasculopathy, Singh [/bib_ref] Moreover, miR200b can inhibit angiogenesis and delay the healing of diabetic wounds by inhibiting the angiogenic factors Ang-1 and ANGPT1. [bib_ref] Maggot excretions/secretions promote diabetic wound angiogenesis via miR18a/19a -TSP-1 axis, Wang [/bib_ref] [bib_ref] Effects of the antagomiRs 15b and 200b on the altered healing pattern..., Pizzino [/bib_ref] These studies suggest that miRNA prevented angiogenesis in diabetic wounds by targeting ANGPT1.
HG level induced decreased endothelial cell activity, migration, and angiogenesis and increased levels of lactate hydrogenase, ROS, and cell apoptosis. [bib_ref] MiR-155 targets PTCH1 to mediate endothelial progenitor cell dysfunction caused by high..., Gao [/bib_ref] miRNAs can inhibit the proliferation and migration of endothelial cells, promote apoptosis, and hinder angiogenesis in diabetic wounds. Pro-inflammatory stress increases the secretion of miR-191 or miR-200b in endothelial cells, and miR-191 targets the downregulation of zonulaoccludans-1 (zo-1), which ultimately affects migration of dermal endothelial cells or fibroblasts in patients with diabetes. [bib_ref] Impairment of Wound Healing in Patients With Type 2 Diabetes Mellitus Influences..., Dangwal [/bib_ref] miR-135a-3p, an anti-angiogenic miRNA, inhibited the proliferation, migration, and tube-like formation of endothelial cells in the matrix gel. miR-135a-3p effectively regulated the p38 signal stimulated by VEGF by targeting HIP1, thus playing an anti-angiogenic role. [bib_ref] MicroRNA-135a-3p regulates angiogenesis and tissue repair by targeting p38 signaling in endothelial..., Icli [/bib_ref] Similarly, miR-155 directly targeted the 3 0 UTR region of PTCH1, inhibited the expression of PTCH1, and mediated endothelial cell dysfunction. [bib_ref] MiR-155 targets PTCH1 to mediate endothelial progenitor cell dysfunction caused by high..., Gao [/bib_ref] In other words, these miRNAs impair the function of endothelial cells by inhibiting their proliferation and migration to achieve anti-angiogenesis.
## Regulatory re-epithelialization
Keratinocytes are important cells for re-epithelialization, and dysregulated miRNAs mainly regulate wound epithelialization by affecting their functions. Studies have found that hADSC-exos overexpressing miR-21-5p promoted the migration and proliferation of keratinocytes in vitro. Studies on the wound model of diabetic rats showed that hADSC-exos significantly accelerated wound healing after 15 days of intervention. Histological examination revealed that hADSC-exos not only significantly promoted re-epithelialization but also angiogenesis and collagen remodeling. miR-21-5p promoted the proliferation and migration of keratinocytes and accelerated the healing of diabetic wounds by activating the Wnt/b-catenin-7 signaling pathway. [bib_ref] Engineered Human Adipose Stem-Cell-Derived Exosomes Loaded with miR-21-5p to Promote Diabetic Cutaneous..., Lv [/bib_ref] In addition, Li et al. [bib_ref] miR-5591-5p regulates the effect of ADSCs in repairing diabetic wound via targeting..., Li [/bib_ref] showed that miR-5591-5p reduced ROS production and cell apoptosis in ADSCs by targeting the AGE/AGER/JNK signal axis, increased the survival rate of ADSCs, and promoted wound reepithelialization, accelerating wound healing in diabetics. These results suggest that miRNA is delivered through exos to activate the function of keratinocytes and promote re-epithelialization of skin wounds.
Matrix metallopeptidases (MMPs) play an indispensable role in the movement of keratinocytes. They regulate the imbalance between the synthesis and degradation of ECM and influence the re-epithelialization and remodeling of wounds. Studies have shown that DFU healing gradually improved with a slow decline in MMP-9 levels and that the severity of DFU is positively correlated with the MMP-9 level. Several studies have shown that Sp1 in keratinocytes can directly bind to the MMP-9 promoter and enhance its expression; however, Sp1 is regulated by miR-129 and miR-335. miR-129 or miR-335 can inhibit the expression of MMP-9 by a targeted reduction of Sp1, thus increasing keratinocyte migration, promoting re-epithelialization, and restoring skin thickness and collagen content. [bib_ref] MicroRNA-129 and -335 Promote Diabetic Wound Healing by Inhibiting Sp1-Mediated MMP-9 Expression, Wang [/bib_ref] That is to say, miRNA activates the function of keratinocytes through targeted regulation of MMP and promotes the re-epithelialization of wounds.
In contrast, other miRNAs inhibit the function of keratinocytes and hinder the re-epithelialization of wounds. miR-203 is the most abundant keratinocyte-specific miRNA in the epidermis and plays an important role in cell differentiation and proliferation. Compared with normal skin tissues, the expression level of miR-203 in patients with DFU was significantly higher and was positively correlated with the severity of DFUs. 14 Studies have shown that miR-203 directly targeted and downregulated IL-8, inhibited the epidermal-mesenchymal transition process, and inhibited keratinocyte proliferation and migration, thereby delaying wound healing. [bib_ref] Acts as an Inhibitor for Epithelial-Mesenchymal Transition Process in Diabetic Foot Ulcers..., Yuan [/bib_ref] Similarly, miR-155 inhibited the expression of FGF7 in diabetic wounds, hindered the proliferation and migration of keratinocytes, and delayed the reepithelialization of wounds. [bib_ref] microRNA-155 inhibition restores Fibroblast Growth Factor 7 expression in diabetic skin and..., Moura [/bib_ref] Additionally, studies have shown that miR-210 inhibited its target gene, E2F3, which is a cell cycle regulator that promotes the G1/S transition, thereby inhibiting keratinocyte proliferation and re-epithelialization. In contrast, anti-miR-210 increased the proliferation and migration of keratinocytes and significantly accelerated wound healing in diabetic mice. 83
## Regulation of ecm remodeling
ECM remodeling is the last stage of wound healing. Compared with non-diabetic skin, both mouse and patients with diabetes skin are biomechanically inferior, with reduced elasticity, decreased maximum stress, decreased collagen content, and a distortion of ECM production and degradation, in which miRNA has a significant influence. [bib_ref] Mechanisms of mesenchymal stem cell correction of the impaired biomechanical properties of..., Zgheib [/bib_ref] [bib_ref] Dysregulation of collagen production in diabetes following recurrent skin injury: contribution to..., Caskey [/bib_ref] Recent studies have shown that MSC treatment can improve the biomechanical properties of damaged diabetic skin by reducing ECM proteolysis. MSC treatment promoted wound healing by upregulating miR-29b, which inhibited the expression of MMP-9 and increased the expression of type I collagen. [bib_ref] Mesenchymal stem cells correct impaired diabetic wound healing by decreasing ECM proteolysis, Xu [/bib_ref] On the contrary, MSCs downregulated the expression of miR-29a, thereby increasing collagen content and correcting the biomechanical properties of diabetic skin. [bib_ref] Mechanisms of mesenchymal stem cell correction of the impaired biomechanical properties of..., Zgheib [/bib_ref] Another study reported that small EVs produced by the pretreatment of HUVECs with AGEs (AGEs-sEVs) inhibited collagen synthesis by activating the autophagy of human skin fibroblasts, thus delaying the wound-healing process in rats. Additionally, miR-106b-5p, a protein enriched in fibroblasts, was upregulated in AGEs-sEVs. miR-106b-5p upregulation by fibroblasts can reduce the expression of ERK1/2, leading to fibroblast autophagy and subsequent collagen degradation. [bib_ref] Endothelial cell-derived small extracellular vesicles suppress cutaneous wound healing through regulating fibroblasts..., Zeng [/bib_ref] Similarly, miR-27-3p inhibited fibroblast function by targeting NOVA1, which resulted in a slower rate of wound healing. [bib_ref] miR-27-3p inhibition restore fibroblasts viability in diabetic wound by targeting NOVA1, Zhang [/bib_ref] Contrarily, anti-miR-378a enhanced the wound-healing process by upregulating synthin-3, which promotes angiogenesis, and vimentin, which promotes the migration and proliferation of fibroblasts. [bib_ref] Anti-microRNA-378a enhances wound healing process by upregulating integrin beta-3 and vimentin, Li [/bib_ref] These miRNAs regulate the synthesis and degradation of collagen by miR-21 is an miRNA with multiple functions, including the regulation of inflammation, angiogenesis, and ECM remodeling during wound healing. miR-21 regulates ECM remodeling by regulating fibroblast proliferation and migration. The expression of miR-21 increases in the late healing process of normal skin wounds but decreases during the healing process of diabetic wounds. 77,105 miR-21 may regulate fibroblast function through the transforming growth factor (TGF)-b1/NF-kB/miR-21 signaling pathway and play a key role in the healing process of diabetic ulcers. [bib_ref] NFkappaB activation is essential for miR-21 induction by TGFb1 in high glucose..., Madhyastha [/bib_ref] In addition, miR-21-3p can also enhance the function of fibroblasts by reducing the expression of its target gene, SPRY1. [bib_ref] MicroRNA-21-3p accelerates diabetic wound healing in mice by downregulating SPRY1, Wu [/bib_ref] Previous studies have found that microvesicles (MVs) extracted from transfected keratinocytes are rich in miR-21, and direct administration of MVmiR-21 significantly promoted the healing of skin wounds in diabetic rats. In-depth studies have found that MVmiR-21 promoted the proliferation, migration, and differentiation of fibroblasts; induced the pro-angiogenesis process of endothelial cells; and mediated proinflammatory responses. Also, MVmiR-21 increased MMP-1 and MMP-3 levels to regulate the migration of fibroblasts by inhibiting tissue inhibitor of metalloproteinases (TIMPs) and upregulating inflammatory factors to enhance the immune response. MVmiR-21 can also downregulate the protein levels of PTEN and reversioninducing-cysteine-rich protein with Kazal motifs, activate the MAPK/ ERK signaling cascade, and induce expression of a-smooth muscle actin (a-SMA) and N-cadherin, which promotes the differentiation of fibroblasts into myofibroblasts. [bib_ref] Human keratinocyte-derived microvesicle miRNA-21 promotes skin wound healing in diabetic rats through..., Li [/bib_ref] miR-21-5p can also regulate keratinocyte function by activating the Wnt/b-catenin/MMP-7 signaling pathway. [bib_ref] Engineered Human Adipose Stem-Cell-Derived Exosomes Loaded with miR-21-5p to Promote Diabetic Cutaneous..., Lv [/bib_ref]
## The regulatory role of lncrna in diabetic wound healing
lncRNAs are a class of ncRNAs longer than 200 nucleotides, which can regulate gene expression at multiple levels including epigenetic regulation, transcription regulation, and post-transcriptional regulation. [bib_ref] Long non-coding RNAs: insights into functions, Mercer [/bib_ref] [bib_ref] Gene regulation by the act of long non-coding RNA transcription, Kornienko [/bib_ref] New evidence shows that lncRNAs have significant expression differences in patients with DFUs and affect wound repair [fig_ref] Table 2: Continued regulating the expression of MMP and can also regulate the function... [/fig_ref].
lncRNA regulation of wound inflammation lncRNAs can play a regulatory role in OS and wound inflammation by controlling the functions of macrophages and T cells. Macrophages polarize from the pro-inflammatory M1 phenotype to the anti-inflammatory M2 phenotype during wound healing. Persistent M1 macrophages may cause chronic inflammation in diabetic wounds. [bib_ref] Exosome-Guided Phenotypic Switch of M1 to M2 Macrophages for Cutaneous Wound Healing, Kim [/bib_ref] The expression of some pro-inflammatory lncRNAs can be induced by hyperglycemia in diabetes mellitus, including lncRNA E330013P06, lncRNA-Gas5, and lncRNA Dnm3os. [bib_ref] Diabetes Mellitus-Induced Long Noncoding RNA Dnm3os Regulates Macrophage Functions and Inflammation via..., Das [/bib_ref] [bib_ref] Long Noncoding RNA GAS5 Regulates Macrophage Polarization and Diabetic Wound Healing, Hu [/bib_ref] [bib_ref] Regulation of inflammatory phenotype in macrophages by a diabetes-induced long noncoding RNA, Reddy [/bib_ref] Because of the regulatory activities of these lncRNAs, the pro-inflammatory M1 macrophages increased, and the anti-inflammatory M2 macrophages decreased, leading to aggravation of wound inflammation. lncRNA E330013P06 was also upregulated in macrophages of diabetic mice and amplified with increased expression of inflammatory genes. Small interfering (si)RNA-mediated E330013P06 gene silencing inhibited the expression of inflammatory genes induced by diabetes stimulation, thereby exerting an anti-inflammatory effect. [bib_ref] Regulation of inflammatory phenotype in macrophages by a diabetes-induced long noncoding RNA, Reddy [/bib_ref] Similar studies have found that lncRNA-Gas5 induced expression of the downstream target gene STAT1, promoted the polarization of macrophages to the pro-inflammatory M1 phenotype, and inhibited the healing of diabetic wounds. [bib_ref] Long Noncoding RNA GAS5 Regulates Macrophage Polarization and Diabetic Wound Healing, Hu [/bib_ref] Furthermore, studies have shown that NF-kB activation induced an increase in the expression of lncRNA Dnm3os under diabetic conditions, aggravating the inflammatory response. RNA fluorescence in situ hybridization showed that lncRNA Dnm3os in macrophages was localized in the nucleus. Overexpression of Dnm3os in macrophages changed the overall modification of histones, upregulated inflammation and immune response genes, and increased phagocytosis. In contrast, RNAi-mediated lncRNA Dnm3os gene silencing attenuated these responses. Additionally, decreased nucleolin levels and increased lncRNA Dnm3os during diabetes facilitated H3K9ac. It may be that the recruitment of histone acetyltransferase resulted in chromatin relaxation, leading to the upregulation of inflammatory target genes and macrophage dysfunction. [bib_ref] Diabetes Mellitus-Induced Long Noncoding RNA Dnm3os Regulates Macrophage Functions and Inflammation via..., Das [/bib_ref] Hence, lncRNA acts on macrophages to regulate their phenotypic conversion and the secretion of inflammatory factors to regulate wound inflammation.
T lymphocytes are the key immune cells involved in wound healing. Their activity is often impaired in diabetic wounds, leading to unbalanced inflammatory reactions. Studies have shown that lncRNA-ENST00000411554 regulated the expression of inflammatory factors in T cell-induced responses by downregulating the downstream target gene MAPK1. However, lncRNA-ENST00000411554 is downregulated in DFU patients, mediating the immune-regulation imbalance of these patients. [bib_ref] Screening and preliminary validation of T lymphocyte immunoregulation-associated long non-coding RNAs in..., Xu [/bib_ref] Additionally, the study reported that the level of lncRNA MALAT1 increased early under the HG level but subsequently decreased, which in turn positively upregulated inflammatory ligands such as SAA3 and ultimately increased the production of inflammatory cytokines such as IL-6 and TNF-a. si-MALAT1 can inhibit ROS production in the early stage of OS in endothelial cells. [bib_ref] Long non-coding RNA MALAT1 regulates hyperglycaemia induced inflammatory process in the endothelial..., Puthanveetil [/bib_ref] Therefore, the expression of lncRNA in the wound is abnormal, leading to dysfunction of T lymphocytes and abnormal inflammation of the wound.
In addition, lncRNA exerts anti-inflammatory effects by inhibiting the activation of inflammatory signaling pathways. In diabetic wounds, the expression of cellular OS-related genes, ROS and NOX2, was upregulated, whereas anti-inflammatory lncRNA Lethe was downregulated. The overexpression of lncRNA Lethe significantly reduces the translocation of p65-NF-kB to the nucleus and reduces expression of NOX2 and the production of ROS, thereby reducing OS and inflammation and accelerating the healing of diabetic wounds. [bib_ref] Long non-coding RNA Lethe regulates hyperglycemia-induced reactive oxygen species production in macrophages, Zgheib [/bib_ref] Similarly, WAKMAR2 inhibits the production of inflammatory chemokines in keratinocytes by regulating the TGFb-WAKMAR2-p65-NF-kB signaling axis. Studies have found that the expression of anti-inflammatory WAKMAR2 is reduced in diabetic wounds, and its expression is induced by TGF-b signals. Knockout of the WAKMAR2 gene can damage the re-epithelialization of human wounds in vitro and increase the level of inflammation; on the contrary, overexpression of WAKMAR2 inhibits the production of inflammatory chemokines by keratinocytes and promotes cell migration. [bib_ref] WAKMAR2, a Long Noncoding RNA Downregulated in Human Chronic Wounds, Modulates Keratinocyte..., Herter [/bib_ref] lncRNA regulates angiogenesis lncRNAs play a role in promoting angiogenesis in diabetic wound repair through regulation of signaling molecules. lncRNA-H19, in which its expression level is significantly reduced in diabetics, is an angiogenesis-promoting lncRNA. It reverses the inhibition of angiogenesis in diabetics by activating the PI3K-Akt pathway. Tao et al. [bib_ref] Extracellular vesicle-mimetic nanovesicles transport LncRNA-H19 as competing endogenous RNA for the treatment..., Tao [/bib_ref] used high-yield EV-like nanoparticles as an effective lncRNA nanodrug delivery system combined with sodium alginate as a carrier to deliver lncRNA-H19 to wounds, and they observed that lncRNA-H19 significantly promoted the regeneration of blood vessels and significantly accelerated the healing of chronic wounds. It has been reported that Nrf2 and HIF-1a can regulate angiogenesis in DFUs by activating MALAT1/HIF-1a signaling. [bib_ref] Role of Nrf2 in MALAT1/ HIF-1a loop on the regulation of angiogenesis..., Jayasuriya [/bib_ref] Expression of lncRNA-MALAT1 was significantly reduced in patients with DFUs, and its reduction was positively correlated with the expression of angiogenic factors Nrf2, HIF-1a, and VEGF. Similar results were observed after knocking out lncRNA-MALAT1. This suggests that Nrf2 can regulate the MALAT1/HIF-1 loop through a positive-feedback mechanism, which may regulate angiogenesis in diabetic wounds. eNOS is a key regulator of endothelial homeostasis and vascular function. Previous studies have shown that lncRNA-Leene can affect endothelial cell function through epigenetic regulation of eNOS expression. In detail, lncRNA-Leene enhancers are located near the eNOS promoters, and lncRNA-Leene promotes the recruitment of RNA Pol II to these promoters to enhance eNOS expression, whereas the inhibition of lncRNA-Leene inhibited the expression of eNOS. [bib_ref] Enhancer-associated long non-coding RNA LEENE regulates endothelial nitric oxide synthase and endothelial..., Miao [/bib_ref] In addition, lncRNAs have anti-angiogenic effects. IGF2 is a key transcription factor that regulates cardiovascular function in patients with diabetes, and its antisense nucleic acid IGF2AS is an lncRNA. Research has shown that the expression of IGF2ASII in myocardial microvascular endothelial cells was upregulated, whereas the expression of IGF2 was downregulated. The inhibition of IGF2AS can upregulate IGF2 and VEGF, enhance the proliferation and invasive ability of endothelial cells, and promote angiogenesis. [bib_ref] Inhibition of long noncoding RNA IGF2AS promotes angiogenesis in type 2 diabetes, Zhao [/bib_ref] lncRNA regulates ECM remodeling
The interaction and abnormal expression of lncRNAs and miRNAs may explain inhibition of ECM remodeling in diabetic wounds. Previous studies have confirmed that the lncRNA-H19/miR-152-3p/PTEN axis can regulate the biological activity and inflammatory response of fibroblasts and influence healing of DFUs. When MSC-exos overexpressing lncRNA-H19 were co-cultured with fibroblasts, their proliferation and migration significantly improved, apoptosis and inflammation were inhibited, and lncRNA-H19 significantly facilitated wound healing in DFU mice. [bib_ref] The MSC-Derived Exosomal lncRNA H19 Promotes Wound Healing in Diabetic Foot Ulcers..., Li [/bib_ref] In addition, H19 can recruit enhancer of zeste (EZH) 2 to mediate HIF-1 histone H3K4me3 methylation, increase the expression of HIF-1a, and enhance the function of fibroblasts. [bib_ref] Autologous blood transfusion augments impaired wound healing in diabetic mice by enhancing..., Guo [/bib_ref] Similarly, lncRNA-MALAT1 facilitated the activation of fibroblasts in diabetic mice by activating the HIF-1a signaling pathway, which increased wound collagen synthesis and facilitated healing. [bib_ref] lncRNA MALAT1 Accelerates Wound Healing of Diabetic Mice Transfused with Modified Autologous..., Liu [/bib_ref]
## Regulatory role of circrnas in diabetic wounds
Wang et al. [bib_ref] Circular RNA hsa_circ_0084443 Is Upregulated in Diabetic Foot Ulcer and Modulates Keratinocyte..., Wang [/bib_ref] reported that 115 circRNAs were upregulated and 111 circRNAs were downregulated in DFUs, suggesting that circRNAs may be involved in the repair process, including cell proliferation and migration, apoptosis and autophagy, and angiogenesis [fig_ref] Table 3: Regulatory role of circRNAs in wound repair [/fig_ref]. [bib_ref] Circular RNA hsa_circ_0084443 Is Upregulated in Diabetic Foot Ulcer and Modulates Keratinocyte..., Wang [/bib_ref] [bib_ref] High glucose-induced circHIPK3 downregulation mediates endothelial cell injury, Cao [/bib_ref] [bib_ref] The Circular RNA Interacts with STAT3, Increasing Its Nuclear Translocation and Wound..., Yang [/bib_ref] [bib_ref] Guidance of circular RNAs to proteins' behavior as binding partners, Luo [/bib_ref] [bib_ref] Profiling and functional analysis of differentially expressed circular RNAs in high glucose-induced..., Jin [/bib_ref] It was reported that the circRNA expression level in HUVECs could be altered by HG levels. Sequencing results showed 214 differentially expressed circRNAs, among which three showed the highest upregulation (hsa_circ_0008360, hsa_circ_0000109, and hsa_circ_0002317). Bioinformatics analysis indicated that these circRNAs regulate the expression of vascular endothelial function and angiogenesis-related genes by targeting miRNAs. [bib_ref] Profiling and functional analysis of differentially expressed circular RNAs in high glucose-induced..., Jin [/bib_ref] HG levels trigger the regulation of endothelial cell damage through the circHIPK3/miR-124/SphK1/ STAT3 axis. In other words, overexpression of circHIPK3 can inhibit HG-induced cell death and apoptosis; contrarily, targeted siRNA silencing of circHIPK3 caused the death of endothelial cells. Further studies have found that HG-induced downregulation of circHIPK3 may lead to the accumulation of miR-124 in HUVECs and human arterial endothelial cells (HAECs), whereas miR-124 inhibits endothelial cell activity, promotes apoptosis, and impairs endothelial cell migration and tube formation. [bib_ref] High glucose-induced circHIPK3 downregulation mediates endothelial cell injury, Cao [/bib_ref] Additionally, studies have shown that ADSC-exo containing mmu_circ_0000250 promoted the expression of its downstream gene SIRT1 by adsorbing miR-128-3p facilitating the activation of autophagy under HG conditions. In vivo experiments showed that exos containing high concentrations of mmu_circ_0000250 accelerated the healing of diabetic wounds. Histopathological examinations show that it inhibited cell apoptosis through the activation of autophagy and facilitated angiogenesis in the skin wound. [bib_ref] Exosomes derived from mmu_circ_0000250-modified adipose-derived mesenchymal stem cells promote wound healing in..., Shi [/bib_ref] circRNAs also play an important role in wound healing by regulating the functions of keratinocytes and fibroblasts. Wang et al. [bib_ref] Circular RNA hsa_circ_0084443 Is Upregulated in Diabetic Foot Ulcer and Modulates Keratinocyte..., Wang [/bib_ref] showed that the circRNA hsa_circ_0084443 upregulated keratinocyte migration and proliferation in DFUs. Zhang et al. [bib_ref] Circ_0075932 in adipocytederived exosomes induces inflammation and apoptosis in human dermal keratinocytes..., Zhang [/bib_ref] found that circ_0075932 in ADSC-EXO induced inflammation and apoptosis of human dermal keratinocytes by directly binding to PUM2 and promoting PUM2-mediated AuroraA/NF-kB pathway activation. Additionally, Yang et al. [bib_ref] The Circular RNA Interacts with STAT3, Increasing Its Nuclear Translocation and Wound..., Yang [/bib_ref] reported that circRNAmotl1 encouraged wound repair by promoting fibroblast proliferation, survival, adhesion, and migration. Research has shown that circRNAmotl1 carries out the above-mentioned activities by promoting the nuclear translocation of STAT3, which binds the DNMT3A promoter to facilitate its transcription and translation. DNMT3A induces the methylation of the miR-17 promoter and reduces the expression of miR-17-5p [fig_ref] Table 4: Regulatory effects of different ncRNAs on key genes and signal pathways [/fig_ref].
## The application prospects of ncrna in diagnosis and treatment
ncRNAs often respond differently in various pathological conditions, and changes in their expression directly reflect the physiological state of cells. Moreover, differences in the expression of ncRNA have been reported in both type 1 diabetes and T2D of animal studies, [bib_ref] MicroRNA miR-27b rescues bone marrow-derived angiogenic cell function and accelerates wound healing..., Wang [/bib_ref] [bib_ref] Identification of Specific miRNAs in Neutrophils of Type 2 Diabetic Mice: Overexpression..., Umehara [/bib_ref] [bib_ref] MicroRNA-132 with Therapeutic Potential in Chronic Wounds, Li [/bib_ref] [bib_ref] The role of microRNA-146a in the pathogenesis of the diabetic wound-healing impairment:..., Xu [/bib_ref] indicating that ncRNA influences both types of diabetic wounds; however, whether insulin resistance is a differentiating factor in T2D wounds remains unknown. miRNA expression profile analysis of fibroblasts extracted from DFUs showed significant distortion that affected important stages of wound repair. [bib_ref] Integrative analysis of miRNA and mRNA paired expression profiling of primary fibroblast..., Liang [/bib_ref] [bib_ref] MicroRNA profiling in cutaneous wounds of diabetic rats, Liu [/bib_ref] Another study identified 58 upregulated lncRNAs and 42 downregulated lncRNAs in DFUs compared with non-patients with diabetes. [bib_ref] Co-expression network analysis revealing the key lncRNAs in diabetic foot ulcers, Yu [/bib_ref] Similarly, HG induced changes in the expression of lncRNAs and related mRNAs. [bib_ref] Detection and analysis of angiogenesis pathway-associated lncRNA expression profiles in human skin..., Tu [/bib_ref] Additionally, there were significant differences in the expression of circRNAs in DFUs, [bib_ref] Circular RNA hsa_circ_0084443 Is Upregulated in Diabetic Foot Ulcer and Modulates Keratinocyte..., Wang [/bib_ref] and the expression profile of circRNAs in HUVECs induced by HG was also significantly different. Further bioinformatics analysis showed that these circRNAs regulated the expression of vascular endothelial function and angiogenesis-related genes by targeting miRNAs. [bib_ref] Profiling and functional analysis of differentially expressed circular RNAs in high glucose-induced..., Jin [/bib_ref] Therefore, we can diagnose and treat diseases by detecting the content of specific ncRNAs.
## Ncrna as a diagnostic biomarker
Previous studies have confirmed that ncRNAs are differentially expressed in patients with diabetes and are closely related to the occurrence and development of the disease. [bib_ref] Co-expression network analysis revealing the key lncRNAs in diabetic foot ulcers, Yu [/bib_ref] [bib_ref] Exploring microRNAs in diabetic chronic cutaneous ulcers: Regulatory mechanisms and therapeutic potential, Nie [/bib_ref] [bib_ref] Relationships of Non-coding RNA with diabetes and depression, An [/bib_ref] Lin et al. [bib_ref] Expression of miR-217 and HIF-1a/VEGF pathway in patients with diabetic foot ulcer..., Lin [/bib_ref] showed that the serum levels of miR-217 in DFU patients was significantly upregulated compared with those of healthy individuals. Moreover, as the Wagner grade increases, the serum miR-217 level gradually increases. In other words, the expression level of mir-217 is positively correlated with the severity of DFUs. Another study also found that compared with normal skin tissue, the expression level of miR-203 in the skin of DFUs was significantly higher, and the expression profile of miR-203 was positively correlated with the severity of DFUs. Moreover, compared with other parameters used to assess the severity of DFUs, the measurement of miR-203 is more accurate and effective. [bib_ref] Quantification of the differential expression levels of microRNA-203 in different degrees of..., Liu [/bib_ref] In addition, studies have found that baseline serum miR-15a and miR-16 levels were positively correlated with restenosis after amputation in T2D patients. [bib_ref] MicroRNA-15a and microRNA-16 impair human circulating proangiogenic cell functions and are increased..., Spinetti [/bib_ref] Therefore, the level of expression of ncRNA may provide a new reference index for the clinical classification DFUs.
In addition, the physiological and pathological changes of the human body may be reflected by the changes in the expression of ncRNA in the blood. For example, circulating exosomal miR-20b-5p has been proven to be an effective and non-invasive biomarker for the early diagnosis of non-small cell lung cancer and nasopharyngeal carcinomas. [bib_ref] Identification of a 7-microRNA signature in plasma as promising biomarker for nasopharyngeal..., Zhang [/bib_ref] [bib_ref] Circulating serum exosomal miR-20b-5p and miR-3187-5p as efficient diagnostic biomarkers for early-stage..., Zhang [/bib_ref] The correlation between miR-20b-5p and diabetes is not yet clear. Further studies have found that miR-20b-5p from circulating exos in patients with diabetes was significantly upregulated, and such exosomal intervention may hinder wound healing. [bib_ref] Circulating Exosomal miR-20b-5p Inhibition Restores Wnt9b Signaling and Reverses Diabetes-Associated Impaired Wound..., Xiong [/bib_ref] Similarly, Dangwal et al. [bib_ref] Impairment of Wound Healing in Patients With Type 2 Diabetes Mellitus Influences..., Dangwal [/bib_ref] reported reduced levels of miR-191 and miR-200b in blood in T2D patients compared with healthy individuals. Patients with diabetes subjects with associated peripheral artery disease and chronic wounds showed higher levels of circulating C-reactive protein and pro-inflammatory cytokines. These results indicate that monitoring the expression of ncRNAs may be adopted as an accurate and effective biomarker for the clinical evaluation of the severity of diabetic wounds.
ncRNA can be used as a target for wound-repair treatment
Recent studies have shown that ncRNAs play important roles in cancer and cardiovascular disease and have indicated most drugs currently have potential ncRNA binding sites.This provides a basis for the development of drugs that use ncRNAs as a therapeutic target. For example, ginsenoside uses miR-23a as a drug target. Downregulating the expression of miR-23a reduced its inhibitory effect on downstream gene IRF-1 and upregulated iNOS, promoted wound angiogenesis, and facilitated the healing of DFUs.Sawaya et al. [bib_ref] Topical mevastatin promotes wound healing by inhibiting the transcription factor c-Myc via..., Sawaya [/bib_ref] reported that the topical application of mevastatin accelerated wound healing; one of the possible mechanisms is that mevastatin induced the expression of lncRNA-Gas5. lncRNA-Gas5 inhibits the expression of the transcription factor c-Myc, which regulates the proliferation and migration of keratinocytes, thereby promoting wound healing.
Inhibition of the healing process in diabetic wounds may partly be due to abnormal expression of ncRNAs. The restoration of the expression level by exogenous supplementation or inhibition of the related ncRNAs may be an ideal method to speed healing of diabetic wounds. miR-21 expression is significantly reduced in diabetic wounds. Lv et al. [bib_ref] Engineered Human Adipose Stem-Cell-Derived Exosomes Loaded with miR-21-5p to Promote Diabetic Cutaneous..., Lv [/bib_ref] loaded exogenous miR-21-5p into hADSC-exos using electroporation and significantly accelerated wound healing. Similarly, Tao et al. [bib_ref] Extracellular vesicle-mimetic nanovesicles transport LncRNA-H19 as competing endogenous RNA for the treatment..., Tao [/bib_ref] delivered lncRNA-H19 to wounds using extracellular-simulated vesicle nanocapsules as the delivery system and observed a significant stimulation of angiogenesis. In addition, miR-146a, which possess anti-inflammatory properties, was downregulated in diabetic wounds. Zgheib et al. [bib_ref] Use of Cerium Oxide Nanoparticles Conjugated with MicroRNA-146a to Correct the Diabetic..., Zgheib [/bib_ref] used cerium oxide nanoparticles to deliver miR-146a to diabetic mice skin wounds, which improved wound inflammation, increased angiogenesis, and significantly accelerated wound closure.
In addition to restoring the expression level of ncRNAs in wounds through a delivery system, it has also been shown that damage resulting from abnormal ncRNA expression can be reversed by directly using antisense ncRNAs or antagonists. MRG-110, a synthetic antiangiogenic molecule, inhibited the expression of miR-92a and accelerated angiogenesis and the healing of both diabetic and non-diabetic wounds. [bib_ref] A synthetic microRNA-92a inhibitor (MRG-110) accelerates angiogenesis and wound healing in diabetic..., Gallant-Behm [/bib_ref] Similarly, the direct use of light-induced anti-miR-92a also promoted skin repair in healing impaired diabetic mice. [bib_ref] Light-inducible antimiR-92a as a therapeutic strategy to promote skin repair in healing-impaired..., Lucas [/bib_ref] Direct application of miR-155 inhibitors to wounds also reduced inflammation, promoted cell migration, and shortened diabetic wound closure time. [bib_ref] MicroRNA-155 Inhibition Promoted Wound Healing in Diabetic Rats, Ye [/bib_ref] [bib_ref] microRNA-155 inhibition restores Fibroblast Growth Factor 7 expression in diabetic skin and..., Moura [/bib_ref] Furthermore, Pizzino et al. [bib_ref] Effects of the antagomiRs 15b and 200b on the altered healing pattern..., Pizzino [/bib_ref] injected antagomir-15b and -200b into the wound edge of the back skin of diabetic mice, which changed the expression level of downstream angiogenesis genes and improved the healing outcome.
Finally, ncRNAs can also be used as therapeutic targets for stem cell therapy to regulate the healing process of diabetic wounds. MSC treatment of diabetic wounds is mainly through the upregulation of the expression level of miR-146a, which is involved in the regulation of immune and inflammatory responses, resulting in downregulation of downstream inflammatory target genes. [bib_ref] The role of microRNA-146a in the pathogenesis of the diabetic wound-healing impairment:..., Xu [/bib_ref] MSCs can also reduce the level of miR-29a, increase the content of collagen, and correct the damaged biomechanical properties of patients with diabetes skin. [bib_ref] Mechanisms of mesenchymal stem cell correction of the impaired biomechanical properties of..., Zgheib [/bib_ref] Similarly, MSCs can inhibit the expression of MMP-9 through the action of miR-29b and increase the expression of type I collagen, promoting wound healing. [bib_ref] Mesenchymal stem cells correct impaired diabetic wound healing by decreasing ECM proteolysis, Xu [/bib_ref] Challenges in the application of ncRNAs
Since ncRNAs have different effects in different stages of diabetic wound healing by regulating gene expression at the molecular level, they have broad application prospects in diabetic wound diagnosis and treatment. However, the current research on ncRNAs has just started, especially lncRNAs and circRNAs, still facing some of the following challenges, hence further research into their clinical application is needed.
First, use of ncRNAs as diagnostic markers due to their insufficient specificity, low sensitivity, and individual differences further limits application as therapeutic targets. ncRNAs may participate in several important processes of wound healing; however, their role may differ at each stage. ncRNAs are dynamic and unstable, differ largely in their mode of action, and possess limited specificity. For example, miR-21 is significantly upregulated during M1 polarization of macrophages, and consequent upregulation of inflammation-related factors exerts a pro-inflammatory effect. [bib_ref] Role of microRNA-21 and Its Underlying Mechanisms in Inflammatory Responses in Diabetic..., Liechty [/bib_ref] [bib_ref] Human keratinocyte-derived microvesicle miRNA-21 promotes skin wound healing in diabetic rats through..., Li [/bib_ref] The expression of miR-21 decreases in the mid-stage of wound healing, which weakens its pro-inflammatory effect and exerts an anti-inflammatory effect; however, it is upregulated in the later stage of wound healing to facilitate the proliferation and migration of keratinocytes and fibroblasts. miR-21 can also target and regulate the expression of downstream genes MMP-1, MMP-3, TIMP3, and TIMP4; regulate the remodeling and degradation of ECM; and promote re-epithelialization [bib_ref] Engineered Human Adipose Stem-Cell-Derived Exosomes Loaded with miR-21-5p to Promote Diabetic Cutaneous..., Lv [/bib_ref] [bib_ref] Role of microRNA-21 and Its Underlying Mechanisms in Inflammatory Responses in Diabetic..., Liechty [/bib_ref] [bib_ref] Human keratinocyte-derived microvesicle miRNA-21 promotes skin wound healing in diabetic rats through..., Li [/bib_ref] [bib_ref] Human fibrocyte-derived exosomes accelerate wound healing in genetically diabetic mice, Geiger [/bib_ref] and can also regulate the transdifferentiation of cells in wound healing. [bib_ref] Direct conversion of injury-site myeloid cells to fibroblast-like cells of granulation tissue, Sinha [/bib_ref] Furthermore, the levels of various ncRNAs are differently affected by the expression level of cytokines or inflammatory factors in the wound, which cause individual differences. For example, TGF-b1 can induce the expression of miR-21 and WAKMAR2, 20,37 and the expression levels of miR-191 and miR-200b are significantly correlated with IL-1b, platelet-derived growth factor (PDGF)-BB, macrophage migration inhibitory factor (MIF), and C-reactive protein. [bib_ref] Impairment of Wound Healing in Patients With Type 2 Diabetes Mellitus Influences..., Dangwal [/bib_ref] Moreover, TNF-a can also induce an increase in the expression of miR-200b during chronic inflammation in diabetes. [bib_ref] Downregulation of endothelial microRNA-200b supports cutaneous wound angiogenesis by desilencing GATA binding..., Chan [/bib_ref] In other words, differences in the level of inflammation in each patient will also lead to differences in the activities of ncRNAs.
On the other hand, in addition to the dynamic and individual differences, the activities of ncRNAs may overlap, and their interactions are complex; that is, ncRNAs have multiple downstream target genes, regulate different signaling pathways, and participate in different physiological processes of various cells and tissues, so their actions differ across cells, tissues, and target sites. [bib_ref] Autologous blood transfusion augments impaired wound healing in diabetic mice by enhancing..., Guo [/bib_ref] [bib_ref] The MSC-Derived Exosomal lncRNA H19 Promotes Wound Healing in Diabetic Foot Ulcers..., Li [/bib_ref] [bib_ref] miRNA-221-3p in Endothelial Progenitor Cell-Derived Exosomes Accelerates Skin Wound Healing in Diabetic..., Xu [/bib_ref] [bib_ref] Staphylococcus aureus Triggers Induction of miR-15B-5P to Diminish DNA Repair and Deregulate..., Ramirez [/bib_ref] [bib_ref] Expression of TGF-b1 and miRNA-145 in patients with diabetic foot ulcers, Zhang [/bib_ref] A disease may be accompanied by several changes in the expression of different ncRNAs, [bib_ref] MicroRNA profiling in cutaneous wounds of diabetic rats, Liu [/bib_ref] [bib_ref] Differentially expressed circulating microRNAs in the development of acute diabetic Charcot foot, Pasquier [/bib_ref] and their functions differ markedly. [bib_ref] miR-23c regulates wound healing by targeting stromal cell-derived factor-1a (SDF-1a/CXCL12) among patients..., Amin [/bib_ref] The expression dynamics and individual differences of ncRNAs as well as their mutual relationship are like a complex network, which greatly increases the challenge of clinical application of ncRNAs. Use of ncRNAs as a diagnostic biomarker is based on changes in their level of expression or restoration of expression through artificial supplementation for treatment purposes. There is a need for a patient-based approach; that is, more information about the patient is needed for the successful application of ncRNAs in disease diagnosis and treatment. In addition, because each ncRNA can regulate different signaling pathways and play different roles in different tissues and cells, research regarding the clinical effectiveness of ncRNAs is lacking. Researchers must also ascertain whether reversing expression levels will have serious side effects, due to individual differences in ncRNAs expression. Moreover, the statistical methods used in various studies may differ. Therefore, an objective and rational comprehensive analysis of the expression differences of these ncRNAs is warranted.
The use of ncRNAs as a target for diabetic wound treatment requires choosing appropriate methods to synthesize the necessary ncRNA mimics or inhibitors, but we currently lack reliable gene-modification technology. There are two treatment strategies: (1) use of ncRNA mimics or viral vectors to trigger the overexpression of previously inhibited ncRNAs or (2) use of chemically modified antisense ncRNA oligonucleotides to inhibit the expression of previously overexpressed ncRNAs. [bib_ref] MicroRNA-155 Inhibition Promoted Wound Healing in Diabetic Rats, Ye [/bib_ref] [bib_ref] MicroRNA-132 with Therapeutic Potential in Chronic Wounds, Li [/bib_ref] [bib_ref] MicroRNA-132 enhances transition from inflammation to proliferation during wound healing, Li [/bib_ref] [bib_ref] Diabetes-induced glucolipotoxicity impairs wound healing ability of adipose-derived stem cells-through the miR-1248/CITED2/HIF-1a..., Xiao [/bib_ref] Plasmids are commonly used to construct overexpression vectors, and lentiviral transfection techniques are used to synthesize ncRNA. These techniques are not suitable for clinical treatment due to biological safety concerns, [bib_ref] Absence of Replication-Competent Lentivirus in the Clinic: Analysis of Infused T Cell..., Cornetta [/bib_ref] which greatly limits their clinical application.
The clinical application of ncRNAs must depend on a safe, efficient, and reliable delivery system, which we currently lack. ncRNA use mainly includes the direct application of synthetic inhibitors or mimics of corresponding ncRNAs. However, ncRNAs are very easily hydrolyzed in the microenvironment of wounds, and it is difficult for free ncRNAs to enter the cell membrane, resulting in little or no effect. [bib_ref] Engineered Human Adipose Stem-Cell-Derived Exosomes Loaded with miR-21-5p to Promote Diabetic Cutaneous..., Lv [/bib_ref] Existing studies have shown that exos are also promising new drug-delivery vehicles because they can protect their payload from chemical and enzymatic degradation and can evade recognition and subsequent elimination by the immune system. [bib_ref] Circulating Exosomal miR-20b-5p Inhibition Restores Wnt9b Signaling and Reverses Diabetes-Associated Impaired Wound..., Xiong [/bib_ref] Several studies have used exo carriers to deliver ncRNAs to wound sites, which can restore the expression level of ncRNAs in the body, regulate immune responses, inhibit inflammation, and promote cell proliferation and angiogenesis. The combination of ncRNAs and exos significantly facilitates and improves the quality of wound healing. [bib_ref] The MSC-Derived Exosomal lncRNA H19 Promotes Wound Healing in Diabetic Foot Ulcers..., Li [/bib_ref] [bib_ref] LPS-preconditioned mesenchymal stromal cells modify macrophage polarization for resolution of chronic inflammation..., Ti [/bib_ref] [bib_ref] Engineered Human Adipose Stem-Cell-Derived Exosomes Loaded with miR-21-5p to Promote Diabetic Cutaneous..., Lv [/bib_ref] [bib_ref] Human fibrocyte-derived exosomes accelerate wound healing in genetically diabetic mice, Geiger [/bib_ref] However, exos have disadvantages, such as complex extraction, low yield, and high cost, hindering clinical application. [bib_ref] Prospective application of exosomes derived from adipose-derived stem cells in skin wound..., Qiu [/bib_ref] Moreover, there is a lack of appropriate technology for the transfer of ncRNAs into exos.
Furthermore, the current research on ncRNAs and their possible application in the treatment of DFUs is basic, and clinical research is still lacking. Moreover, the models used in basic research possess certain limitations. Existing in vivo studies are mainly restricted to mouse and rat wound models. Wound healing in rats is mainly completed by skin contraction, whereas in humans, it involves granulation tissue repair; thus, the rat model cannot simulate the physiological repair process of human skin defects. In addition, the existing diabetic wound model is mainly an acute wound model, which cannot simulate the chronic state of complications caused by long-term hyperglycemia. The use of diabetic pigs as a model in the studies of wound healing is an ideal substitute because their skin is structurally similar to human skin. [bib_ref] Topical mevastatin promotes wound healing by inhibiting the transcription factor c-Myc via..., Sawaya [/bib_ref] [bib_ref] Use of Cerium Oxide Nanoparticles Conjugated with MicroRNA-146a to Correct the Diabetic..., Zgheib [/bib_ref] Conclusions ncRNAs, especially miRNAs, are involved in the wound-healing process, including the regulation of inflammation, granulation tissue formation, re-epithelialization, and ECM remodeling. With the advancement in technology and research, ncRNAs are expected to become a new, accurate, and effective biomarker for disease diagnosis and potential therapeutic targets. However, current research on ncRNAs is still in its infancy, especially lncRNA and circRNAs. Moreover, most of these studies are mainly basic research; hence, preclinical and clinical studies are required in order to verify clinical applications. More specific ncRNAs should be screened as diagnostic markers and therapeutic targets. Research and development of more efficient and convenient ncRNA delivery systems are also warranted.
[fig] Figure 1: Pathological characteristics of diabetic wounds (A-C) Diabetic wounds are mainly characterized by (A) chronic inflammation, (B) vascular dysfunction, and (C) delayed re-epithelization and disorder of extracellular matrix remodeling. [/fig]
[fig] Figure 2: Biogenesis and regulation mechanism of miRNA function (1) The miRNA gene was cleaved and processed by RNA polymerase II enzyme and endonuclease Drosha III to form pre-miRNA. The Exportin-5 protein then transports the pre-miRNA to the cytoplasm where it is digested by Dicer to produce mature miRNA, which is loaded on Ago2 to form RISC. Then, it inhibits transcription (A), regulates the degradation of mRNA (B), combine with 5 0 UTR (C) and coding sequence (D) ,etc. (2) During hyperglycemia and inflammation, the effects of p65 and miR-21 promoter are enhanced, whereas miR-21 expression is induced. RNase IRE1alpha reduces the levels of miR-200 and miR-466 to that of pre-miRNA. In addition, hyperglycemia can change the expression of miRNA epigenetics by affecting the DNMT methylation process. The expression-imbalanced miRNA can inhibit the expression of the target gene by interacting with the 3 0 untranslated region (3 0 UTR) of the target mRNA, inducing mRNA degradation, and inhibiting translation. In addition, miRNAs can also interact with 5 0 UTR and gene promoters and induce methylation modification. [/fig]
[fig] Figure 3: Overview of the mechanisms of non-coding RNA (ncRNA) function lncRNAs can play a regulatory role in multiple levels of epigenetics (A), transcription (B), and post-transcription (C,D). HATs, histone acetyl transferases; EZH, enhancer of zeste; TF, transcription factor. [/fig]
[fig] Figure 4: Mechanisms of circRNA function circRNAs can regulate gene transcription mediated by STAT3 nuclear translocation (A) and act as competitive endogenous RNAs to mitigate the inhibitory effect of miR-124, miR-128-3p, and other miRs on the downstream target genes to regulate angiogenesis (B) [/fig]
[fig] Figure 5: Regulatory effects of certain common ncRNAs on the pathology of diabetic wounds The diabetic wound microenvironment induces ncRNA changes, which in turn, causes an imbalance in the downstream gene expression, thereby impairing the functions of immune cells, endothelial cells, keratinocytes, and fibroblasts in the wound and causing cytokine imbalance, further exacerbating the diabetic wound. Moreover, these three pathological processes are inter-related. [/fig]
[table] Table 1: Functions and mechanisms of miRNAs in diabetic wounds [/table]
[table] Table 3: Regulatory role of circRNAs in wound repair [/table]
[table] Table 4: Regulatory effects of different ncRNAs on key genes and signal pathways [/table]
|
Circulating Total Glutathione in Normal Tension Glaucoma Patients: Comparison with Normal Control Subjects
[bib_ref] Glaucomatous outf low pathway and oxidative stress, Sacca [/bib_ref] [bib_ref] Age-related nuclear cataract-oxidation is the key, Truscott [/bib_ref] [bib_ref] Changes in antioxidant enzyme activity and malondialdehyde level in patients with age-related..., Yildirim [/bib_ref] [bib_ref] A new clue to glaucoma pathogenesis, Chen [/bib_ref] [bib_ref] Glutathione metabolism and oxidative stress in neurodegeneration, Dringen [/bib_ref] [bib_ref] Review: the role of glutathione in the regulation of apoptosis, Hall [/bib_ref] [bib_ref] Physiologic neutralization mechanisms and the response of the corneal endothelium to hydrogen..., Riley [/bib_ref] [bib_ref] Roles of catalase and the glutathione redox cycle in the regulation of..., Costarides [/bib_ref] [bib_ref] Oxidative stress markers in aqueous humor of glaucoma patients, Ferreira [/bib_ref] [bib_ref] Serum autoantibody against glutathione S-transferase in patients with glaucoma, Yang [/bib_ref] [bib_ref] Glutathione S transferase M1 and T1 genetic polymorphisms are related to the..., Unal [/bib_ref] [bib_ref] Systemic reduction in glutathione levels occurs in patients with primary open-angle glaucoma, Gherghel [/bib_ref] [bib_ref] Glutathione depletion induces differential apoptosis in cells of mouse retina, in vivo, Roh [/bib_ref] [bib_ref] Differential expression of heat shock protein mRNAs under in vivo glutathione depletion..., Park [/bib_ref] [bib_ref] The prevalence of glaucoma in Korean careermen, Choe [/bib_ref] [bib_ref] The prevalence of primary open-angle glaucoma in Japanese: the Tajimi Study, Iwase [/bib_ref] [bib_ref] Prevalence of primary open-angle glaucoma in central South Korea the Namil study, Kim [/bib_ref] [bib_ref] Prevalence of primary open-angle glaucoma in central South Korea the Namil study, Kim [/bib_ref] [bib_ref] Vascular dysregulation: a principal risk factor for glaucomatous damage?, Flammer [/bib_ref] [bib_ref] What is the present pathogenetic concept of glaucomatous optic neuropathy?, Flammer [/bib_ref] [bib_ref] Delayed functional loss in glaucoma: LII Edward Jackson Memorial Lecture, Brubaker [/bib_ref]
# Materials and methods
## Study sample
Patients with NTG attending the glaucoma clinic of the Catholic Eye Center at St. Mary's Hospital (Seoul, Korea) between June 2008 and October 2008 were considered for inclusion in this prospective study. Ethical approval was obtained from the committee of St. Mary's Hospital institutional review board, and written informed consent was received from all subjects before entry into the study.
Patients were diagnosed as having NTG if IOP measurements were less than 22 mmHg by Goldmann applanation tonometry, characteristic glaucomatous cupping of the optic disc was observed on funduscopic examination, normal open anterior chamber angles was seen on gonioscopy, and repeatable visual field defects consistent with the diagnosis of glaucoma were observed in the results obtained with program 24-2 of the Humphrey field analyzer (Carl Zeiss Meditec, Dublin, CA, USA). The study's ocular exclusion criteria included any IOP measurement greater than 21 mmHg, narrow iridocorneal angles, evidence of secondary open-angle glaucoma, history of previous intraocular surgery, and other nonglaucomatous ocular diseases including cataract, diabetic retinopathy (DR) and age-related macular degeneration (ARMD). The control group was recruited from a group of age-and gender-matched volunteers and was composed of subjects who had never had glaucoma or other ocular diseases such as cataract, DR, or ARMD.
Other exclusion criteria for both groups were smoking and a history of any chronic systemic disease with presumed low GSH level, including autoimmune diseases, alcoholic liver disease, cancer, and diabetes mellitus. After this selection, the experimental group was narrowed to 19 patients with NTG, and that control group contained 30 subjects.
## Blood sampling
Subjects were instructed to fast from midnight to 8 a.m. on the morning of the test. All blood samples were obtained by a qualified registered nurse in the morning, between 8 and 10 a.m. Four milliliters of blood were collected in EDTA-treated tubes (to prevent oxidation) by venipuncture to the antecubital vein. One thousand microliters of blood was then transferred into a centrifuge tube for initial processing. Red blood cells (RBC) were separated from plasma by centrifugation at 600 ×g for 10 minutes. The pellet RBCs were then transferred into another microcentrifuge tube and washed twice with 3 volumes of phosphate buffered saline (pH = 7.4). A 200 microliter aliquot of RBC pellets was placed in another microcentrifuge tube, to which 200 microliters of 5% 5-sulfosalicylic acid (SSA) were added and then incubated for 10 minutes at 4˚C. After centrifugation at 10,000 ×g for 10 minutes, 100 microliters of the supernatant were immediately cooled at -70˚C. All of these procedures were performed within 2 hours from the time of blood collection. The next steps of the GSH analysis were performed within 10 days.
# Glutathione analysis
Total GSH levels were assessed by the 5, 5'-dithiobis-(2nitrobenzoic acid) recycling procedure, as described in a previous study [bib_ref] Delayed functional loss in glaucoma: LII Edward Jackson Memorial Lecture, Brubaker [/bib_ref]. The glutathione assay kit from Sigma (Glutathione assay kit, product code CS0260; St Louis, MO, USA) was used.
A standard curve from 3.125 to 50 nanomoles in doublefold increments using a GSH standard solution was prepared for each 96-well plate. The standards contained the same final concentrations of SSA as used for the samples.
To each well of a 96-well plate, 150 microliters of working buffer and 10 microliters of standard or sample were added in duplicate, and the plate was incubated at room temperature for 5 minutes. Finally, 50 microliters of 0.16 mg/mL nicotinamide adenine dinucleotide phosphate (NADPH) was added, and the plate was read at 412 nm using a 96well plate reader. A standard curve was then generated by linear regression.
# Statistical analysis
The statistical analysis was performed on a computer equipped with SPSS ver. 12 (SPSS Inc., Chicago, IL, USA). Data are expressed as the mean ± standard deviation. Differences between groups at baseline for age, mean deviation (MD) and pattern standard deviation (PSD) were calculated using Student's t-test. Difference between groups at baseline for gender was calculated using the χ 2 -test (Fisher's exact test). Differences between the two study groups in blood GSH level were computed by analysis of covariance using age as a cofactor. Differences between males and females in blood GSH level were computed by Student's t-test for each study group. A correlation was performed to determine the association of age and blood GSH level. A multiple regression was performed to test the influences of age, gender, and blood GSH level on MD and PSD from the 24-2 test pattern of the Humphrey field analyzer. Results with p = 0.05 were considered statistically significant.
The minimum sample size needed was calculated from the formula [(u + v)(σ 12 + σ 22 ) / (μ1 -μ2) 2 ] with 90% power and a 5% level of significance. The sample size needed was determined to be 18 for each group [(1.28 + 1.96) 2 (100 2 + 100 2 ) / (300 -200) 2 ] using the reported data [bib_ref] Glutathione depletion induces differential apoptosis in cells of mouse retina, in vivo, Roh [/bib_ref].
# Results
Nineteen patients with NTG and 30 normal control subjects were included in this study. The demographics of the study groups are described in [fig_ref] Table 1: Patient demographics [/fig_ref]. There was no significant difference in age ( p = 0.056) or gender ( p = 0.337, χ 2 = 0.921) distribution between the NTG group and control group. There were significant differences in MD (p = 0.000) and PSD (p = 0.000) between the two groups.
Total glutathione level in circulating blood was 524.02 ± 231.09 nmol and 586.06 ± 156.08 nmol in the NTG group and the control group, respectively. Circulating total GSH levels were not statistically different between the NTG group and control group ( p = 0.121, F = 2.212) [fig_ref] Figure 1: Median and interquartile range values of total circulating glutathione [/fig_ref]. In the NTG group, the mean glutathione level was 589.41 ± 239.64 nmol in males and 451.35 ± 210.39 nmol in females. In the control group, the mean glutathione level was 618.98 ±208.16 nmol in males and 569.61 ± 125.70 nmol in females. There was no statistically significant difference in total GSH level between males and females in the NTG group ( p = 0.424) or in the control group ( p = 0.202) [fig_ref] Figure 3: Median and interquartilic range values of total glutathione [/fig_ref]. There was no significant assotciation be-tween age and total GSH level in the NTG group (p = 0.171, r = -0.328) or in the normal group ( p = 0.380, r = -0.166) . In the NTG group, after correcting for age and gender influences on systemic total GSH level, there was no significant association between GSH level and visual field parameter (MD: p = 0.226, R 2 = 0.260; PSD: p = 0.200, R 2 = 0.275) .
# Discussion
Glutathione is normally present in the cornea, lens, trabecular meshwork, ciliary body, and retina. Changes of glutathione and related enzyme activity in the lens, cornea, retina, and other eye tissues occur with aging, cataract, diabetes, irradiation, and administration of some drugs. Many ocular diseases including keratoconus, cataract, glaucoma, diabetic retinopathy, and age-related macular degeneration may occur by oxidative damage induced by an abnormal glutathione redox state. In terms of glaucoma, the major interest lies on two specific structures, the trabecular meshwork and retina, because the trabecular meshwork is a rate-limiting structure of aqueous outflow pathway controlling intraocular pressure, and the retina is one of the target structures of glaucomatous optic neuropathy.
Studies on the calf trabecular meshwork (TM) have shown the presence of glutathione, glutathione reductase, and glucose-6-phosphate dehydrogenase, indicating that the TM is well supplied with the capacity to generate and use NADPH to maintain GSH. It has been supposed that GSH is able to protect the TM against H 2 O 2 -induced oxidative damage, which would decrease the aqueous humor outflow [bib_ref] Glaucomatous outf low pathway and oxidative stress, Sacca [/bib_ref]. Altered GSH and GSH activities have also been reported in the trabecular meshwork and aqueous humor of patients with glaucoma [bib_ref] Oxidative stress markers in aqueous humor of glaucoma patients, Ferreira [/bib_ref] [bib_ref] Serum autoantibody against glutathione S-transferase in patients with glaucoma, Yang [/bib_ref] [bib_ref] Assay of glutathione, glutathione disulfide, and glutathione mixed disulfides in biological samples, Akerboom [/bib_ref] [bib_ref] Glutathione reductase of calf trabecular meshwork, Nguyen [/bib_ref] [bib_ref] Effect of lipid peroxidation inhibition on retinal ganglion cell death, Levin [/bib_ref] [bib_ref] Oxidative deoxyribonucleic acid damage in the eyes of glaucoma patients, Izzotti [/bib_ref]. All of these results imply that reduced activity of GSH in the TM may influence aqueous outflow facility and thus elevate IOP.
Additional studies of the role of glutathione in glaucoma pathogenesis are related to studies of the trabecular meshwork. However, the retina is a major target organ in glaucoma pathogenesis, and glutathione also exists in the retina. Therefore, abnormal glutathione function within the retina may play an important role in glaucoma. Although the function of glutathione in the retina is not fully understood, there is some evidence that it may have a role in neuroprotection under harmful conditions such as ischemia and oxidative damage. In the retina, glutathione is predominately localized to Muller cells and horizontal cells. According to a study by Schutte and Werner [bib_ref] Redistribution of glutathione in the ischemic rat retina, Schutte [/bib_ref] , using immunocytochemical methods for glutathione, postmortem ischemic injury longer than 10 minutes resulted in strong labeling of neurons, particularly of retinal ganglion cells, whereas Muller cells were essentially devoid of immunoreactivity. This means that glutathione molecules were transferred from Muller cells to retinal ganglion cells to protect ganglion cells from ischemic injury. In a study by Carter-Dawson et al. [bib_ref] Glutathione content is altered in Muller cells of monkey eyes with experimental..., Carter-Dawson [/bib_ref] , Muller cells in glaucomatous monkey retinas showed significantly greater immunoreactivity for glutathione than control retinas, increasing with the duration of elevated intraocular pressure.
According to our previous study, a 15% reduction of total retinal GSH level by intraperitoneal injection of buthionine sulfoximine caused retinal neuronal apoptosis in mice [bib_ref] Glutathione depletion induces differential apoptosis in cells of mouse retina, in vivo, Roh [/bib_ref]. To reach this level of GSH in the retina, circulating total GSH might need to be lowered even further. In the same mouse, total GSH level in the heart was reduced to 60% of that of the control mouse (unpublished data). We hypothesized that systemic total glutathione level might be changed in NTG patients in whom there may be additional factors other than increased IOP for glaucomatous damage. If there is little influence of IOP on glaucomatous optic neuropathy in NTG pathogenesis, and if oxidative damage is one of the major risk factors for NTG, the antioxidant capacity of GSH might have been changed in NTG patients. However, as the results of this study show, there was no difference in circulating GSH level between the NTG group and control group. This difference in results between human and animal subjects is possibly due to the difference in background factors such as genetic homogeneity and diet. Systemic glutathione level can change within a day and is affected by diet and medications. We obtained blood samples at a fixed time (8 a.m.) after fasting, as in the previous report [bib_ref] Systemic reduction in glutathione levels occurs in patients with primary open-angle glaucoma, Gherghel [/bib_ref] , in order to reduce the sampling bias.
Gherghel et al. [bib_ref] Systemic reduction in glutathione levels occurs in patients with primary open-angle glaucoma, Gherghel [/bib_ref] demonstrated that patients with POAG exhibit low levels of circulating glutathione, suggesting a general compromise of the antioxidative defense. In Gherghel's study, total GSH level in the POAG group was about 75% (225.27 ± 83.03) of that of the control group (332. . In addition to the IOP range, there is one more thing we should consider about the difference in results between Gherghel's study and ours. In contrast to Caucasians, Korean and Japanese people have a higher prevalence of NTG than POAG with a 3.5-fold elevation in pressure. In Choe and Hong [bib_ref] The prevalence of glaucoma in Korean careermen, Choe [/bib_ref] 's study of Koreans, the estimated prevalence of POAG in the population over 20 years of age was 2.04%, and the prevalence of NTG was 1.71% (83.8% of POAG). In Iwase et al. [bib_ref] The prevalence of primary open-angle glaucoma in Japanese: the Tajimi Study, Iwase [/bib_ref] 's study of Japanese individuals, the estimated prevalence of POAG in the population older than 40 years of age was 3.9%, and the prevalence of cases of POAG with IOP levels of 21 mmHg or less was 3.6% (92.3% of POAG). Korean and Japanese people might have different risk factors compared to Caucasian populations. Different racial characteristics may lead to different results. These issues illustrate the importance of further studies of POAG with elevated IOP in Korea because of the absence of data on systemic glutathione levels in Korean POAG patients.
Although there were no statistically significant differences in total GSH level between the two groups in this study, the range of total GSH level in the NTG group showed a wider distribution compared to that of the control group. The proportions of patients with very low total GSH level and with very high total GSH level were larger in the NTG group. Considering the smaller sample size of the NTG group, this characteristic might be more meaningful. From this point of view, we can also hypothesize that the homeostasis of systemic glutathione-related antioxidant capacity may be inadequately controlled in the NTG patient. Both a low level and a high level of total GSH may be related to insufficient anti-oxidant activity against reactive oxygen species.
It is known that blood GSH level is influenced by gender, with men demonstrating higher levels than women. Our results found that mean total GSH level was slightly higher in men then in women (NTG group, 589.41 ± 239.64 nmol in males and 451.35 ± 210.39 nmol in females; control group, 618.98 ± 208.16 nmol in males and 569.61 ± 125.70 nmol in females), but there were no significant differences in blood GSH level between men and women in either the NTG group ( p = 0.424) or the control group ( p = 0.202). This might be due to either the small sample size or racial characteristics. There are few data on normal circulating GSH levels in the Korean population [bib_ref] The change of blood reduced glutathione according to postmenopausal HRT: GSH as..., Kim [/bib_ref] [bib_ref] Glutathione levels in Helicobacter pylori-infected gastric mucosa, Park [/bib_ref]. Further population-based studies are needed to fully explain the results of this study.
Age is known as one of the influencing factors on systemic GSH level. It has been reported that about half of healthy elderly people show low blood GSH level. A negative correlation between age and blood GSH level has also been reported. Our results demonstrated a slightly decreasing trend of GSH level as age increases but found no significant correlation between age and total circulating GSH level in either the NTG group or control group. This might be also due to the small sample size.
We also expected that if the systemic total GSH level is related to the range of glaucomatous optic neuropathy, then the extent of visual field damage in NTG patients might be proportional to the decrement of total GSH level. But the results of this study also showed no significant relation between the MD and PSD of the visual field test and total GSH level. We think this might be due to the finding of no significant difference in total GSH level. Since there was no difference of total GSH level between the NTG group and normal subject group, the extent of visual function may not have been influenced by GSH level. Another possibility is that there might be a change in circulating GSH only in patients with fast progression of optic nerve damage. If we had had a greater number of patients, we could have divided patients into two groups with progressing and non-progressing optic nerve damage and compared them. However, because normal tension glaucoma is a slowly progressing disease in most patients, this would require enormous time and effort to recruit a sufficient number of volunteers with progressive disease.
A major concern of this study is that the study results came from only one measurement of serum glutathione level. Therefore, these results might not be able to represent the long-term status of patients. However, because there are few reports about the relationship between normal tension glaucoma and glutathione, this study represents a meaningful step toward understanding the pathophysiology of normal tension glaucoma.
## Conflict of interest
No potential conflict of interest relevant to this article was reported.
of the authors and do not necessarily represent those of Merck & Co., Inc.
[fig] Figure 1: Median and interquartile range values of total circulating glutathione (GSH) level in the control group and normal tension glaucoma (NTG) group. Total GSH levels were not statistically different between the two groups ( p = 0.121, F = 2.212). [/fig]
[fig] Figure 5, Figure 2, Figure 4: Correlation between total glutathione (GSH) level and age in the control group. There was no significant association between total GSH level and age (r = -0.166, p = 0.380). Median and interquartilic range values of total glutathione (GSH) level in the normal tension glaucoma (NTG) group; comparison between male and female patients. There was no significant difference between males and females ( p = 0.424) in total circulating GSH level. Correlation between total glutathione (GSH) level and age in year in the normal tension glaucoma group. There was no significant association between total GSH level and age (r = -0.328, p = 0.171). [/fig]
[fig] Figure 3: Median and interquartilic range values of total glutathione (GSH) level in the control group; comparison between male and female patients. There was no significant difference between males and females ( p = 0.202) in total circulating GSH level. [/fig]
[fig] Figure 7 0, Figure 6: Correlation between total glutathione (GSH) level and pattern standard deviation (PSD) in the normal tension glaucoma group. After correcting for age and gender influence, there was no significant association between GSH level and PSD ( p = Correlation between total glutathione (GSH) level and mean deviation (MD) in the normal tension glaucoma group. After correcting for age and gender influence, there was no significant association between GSH level and MD ( p = 0.226, R 2 = 0.260). [/fig]
[table] Table 1: Patient demographics (mean ± SD) Statistical analysis was performed with Student's t-test (age, MD, and PSD) and Fisher's exact test (sex). NTG = normal tension glaucoma; MD = mean deviation; PSD = pattern standard deviation. [/table]
|
Effects of Atrazine, Metolachlor, Carbaryl and Chlorothalonil on Benthic Microbes and Their Nutrient Dynamics
Atrazine, metolachlor, carbaryl, and chlorothalonil are detected in streams throughout the U.S. at concentrations that may have adverse effects on benthic microbes. Sediment samples were exposed to these pesticides to quantify responses of ammonium, nitrate, and phosphate uptake by the benthic microbial community. Control uptake rates of sediments had net remineralization of nitrate (21.58 NO 3 mg gdm 21 h 21 ), and net assimilation of phosphate (1.34 PO 4 mg gdm 21 h 21 ) and ammonium (0.03 NH 4 mg gdm 21 h 21 ). Metolachlor decreased ammonium and phosphate uptake. Chlorothalonil decreased nitrate remineralization and phosphate uptake. Nitrate, ammonium, and phosphate uptake rates are more pronounced in the presence of these pesticides due to microbial adaptations to toxicants. Our interpretation of pesticide availability based on their water/solid affinities supports no effects for atrazine and carbaryl, decreasing nitrate remineralization, and phosphate assimilation in response to chlorothalonil. Further, decreased ammonium and phosphate uptake in response to metolachlor is likely due to affinity. Because atrazine target autotrophs, and carbaryl synaptic activity, effects on benthic microbes were not hypothesized, consistent with results. Metolachlor and chlorothalonil (non-specific modes of action) had significant effects on sediment microbial nutrient dynamics. Thus, pesticides with a higher affinity to sediments and/or broad modes of action are likely to affect sediment microbes' nutrient dynamics than pesticides dissolved in water or specific modes of action. Predicted nutrient uptake rates were calculated at mean and peak concentrations of metolachlor and chlorothalonil in freshwaters using polynomial equations generated in this experiment. We concluded that in natural ecosystems, peak chlorothalonil and metolachlor concentrations could affect phosphate and ammonium by decreasing net assimilation, and nitrate uptake rates by decreasing remineralization, relative to mean concentrations of metolachlor and chlorothalonil. Our regression equations can complement models of nitrogen and phosphorus availability in streams to predict potential changes in nutrient dynamics in response to pesticides in freshwaters.
# Introduction
Agricultural activities, such as crop protection via pesticides, are increasing in response to global human population growth (projected to reach 9 billion by 2050; [bib_ref] Smarter pest control. The pesticide paradox. Introduction, Enserink [/bib_ref]. In the last decade, U.S. pesticide sales have increased ,10% with 80% of these pesticides used for agricultural activities [bib_ref] Smarter pest control. The pesticide paradox. Introduction, Enserink [/bib_ref]. The continued growth of the human population, coupled with the need for more efficient agricultural practices, will undoubtedly yield future increases in the occurrence of pesticides in freshwaters. Further, despite decades of research on agricultural pesticides, recent calls to action have highlighted the need to fill critical knowledge gaps in our understanding of how pesticides may adversely affect aquatic ecosystems [bib_ref] Wildlife ecotoxicology of pesticides: can we track effects to the population level..., Köhler [/bib_ref] , [bib_ref] Prioritizing research for trace pollutants and emerging contaminants in the freshwater environment, Murray [/bib_ref].
Once in the aquatic ecosystem, pesticides may have adverse effects on organisms ranging from direct toxicity to indirect effects such as changes in growth or behavior [bib_ref] The effects of pesticides, pH, and predatory stress on amphibians under mesocosm..., Relyea [/bib_ref]. At higher concentra-tions, such as those following spring runoff, direct mortality results across diverse organisms including tadpoles [bib_ref] The effects of pesticides, pH, and predatory stress on amphibians under mesocosm..., Relyea [/bib_ref] , bluegill [bib_ref] Pesticide toxicity index for freshwater aquatic organisms, Munn [/bib_ref] , and benthic organisms (e.g., amphipods and chironomids, [bib_ref] Analyzing effects of pesticides on invertebrate communities in streams, Liess [/bib_ref]. However, at lower concentrations, sub-lethal effects can result in altered respiration rates [bib_ref] Fungicide-induced declines of freshwater biodiversity modify ecosystem functions and services, Mcmahon [/bib_ref] , organismal growth [bib_ref] The effects of pesticides, pH, and predatory stress on amphibians under mesocosm..., Relyea [/bib_ref] and fecundity [bib_ref] Effects of common-use pesticides on developmental and reproductive processes in Daphnia, Kashian [/bib_ref]. In streams, benthic microbes are an important component of aquatic ecosystems and are integral in nutrient and energy dynamics [bib_ref] The role of benthic invertebrate species in freshwater ecosystems: zoobenthic species influence..., Covich [/bib_ref]. For example, nitrate, ammonium and phosphorus are cycled by benthic microbes through assimilation and remineralization [bib_ref] N uptake as a function of concentration in streams, Dodds [/bib_ref]. These processes are influenced by N concentration in freshwaters [bib_ref] Nitrogen Retention, Removal, and Saturation in Lotic Ecosystems, Bernot [/bib_ref] , [bib_ref] Nitrogen saturation in stream ecosystems, Earl [/bib_ref]. At saturated conditions, such as those in agricultural streams with high input of N from fertilizer runoff, more N is available, due to microbial uptake saturation (i.e. biota have reach their N demand; [bib_ref] Nitrogen and phosphorus uptake in two Idaho (USA) headwater wilderness streams, Davis [/bib_ref] or an increase in heterotrophic mineralization [bib_ref] Nitrogen Retention, Removal, and Saturation in Lotic Ecosystems, Bernot [/bib_ref]. Further, at increasing N concentrations, PO 4 often becomes a secondary limiting nutrient [bib_ref] Nitrogen saturation in stream ecosystems, Earl [/bib_ref].
In the Midwestern U.S., two herbicides (atrazine and metolachlor), one insecticide (carbaryl) and one fungicide (chlorothalonil) have both high usage rates and prevalence in receiving waters [bib_ref] Occurrence of pesticides in shallow groundwater of the United States: Initial results..., Kolpin [/bib_ref] [bib_ref] Pesticides in the nation's streams and ground water, Gilliom [/bib_ref]. Atrazine is a triazine herbicide used predominantly in corn production for control of broadleaf and grassy weeds [bib_ref] Relative mobilities of atrazine, five atrazine degradates, metolachlor, and simazine in soils..., Kruger [/bib_ref] with a half-life in water at pH seven of 86 days . Metolachlor is a chloroacetanilide herbicide that inhibits mitosis and cell division . Metolachlor is stable in water at pH seven . Atrazine and metolachlor were detected in U.S. freshwaters at peak concentrations of 201 mg/L and 77.6 mg/L [fig_ref] Table 1: Detection frequency and concentrations of atrazine, metolachlor, carbaryl, and chlorothalonil in U [/fig_ref]. Carbaryl is a carbamate family insecticide that inhibits the enzyme acetylcholinesterase , with a half-life in water of 12 days at pH seven . Chlorothalonil is a fungicide used in U.S. agriculture; it is stable in water at pH seven . Its mode of action is by binding to glutathione and negatively affecting cellular respiration [bib_ref] Fungicide-induced declines of freshwater biodiversity modify ecosystem functions and services, Mcmahon [/bib_ref]. Carbaryl and chlorothalonil were detected in U.S. freshwaters at peak concentrations of 4.8 mg/L and 0.3 mg/L [fig_ref] Table 1: Detection frequency and concentrations of atrazine, metolachlor, carbaryl, and chlorothalonil in U [/fig_ref].
These pesticides are detected in freshwater ecosystems at concentrations that adversely affect biota and human health [fig_ref] Table 2: Toxicity and Octanol-water partition coefficient of atrazine, metolachlor, carbaryl and chlorothalonil to... [/fig_ref]. However, research has focused primarily on the impacts of agricultural pesticides on fish and invertebrates; little is known about how exposure to pesticides may directly influence benthic nutrient dynamics and overall ecosystem function. For example, ecotoxicology studies addressing the effects of pesticides focus primarily on non-benthic vertebrates (e.g., bluegill, [bib_ref] Pesticide toxicity index for freshwater aquatic organisms, Munn [/bib_ref] , and benthic invertebrates (e.g., amphipods and chironomids, [bib_ref] Analyzing effects of pesticides on invertebrate communities in streams, Liess [/bib_ref] with few studies conducted on sediment microbial dynamics [bib_ref] Characterisation of new strains of atrazine-degrading Nocardioides sp. isolated from Japanese riverbed..., Satsuma [/bib_ref]. Benthic microbial communities influence nutrient cycling (e.g. uptake, remineralization) by affecting fluxes as consumers or sources. Thus, benthic microbial communities are an important component of the freshwater ecosystem [bib_ref] The influence of ammonium, nitrate, and dissolved oxygen concentrations on uptake, nitrification,..., Kemp [/bib_ref]. These nutrient dynamics are affected by the presence of pesticides. Specifically, pesticides can reduce microbial activity that contributes to nutrient cycling (e.g. Volvox spp., Botryococcus spp., Synedra spp.) [bib_ref] The impact of agricultural runoff on stream benthos in Hong Kong, Neumann [/bib_ref] , and change species composition by favoring microbes with enhanced pesticide degradation capacities. Also, pesticides can become nutrient sources by providing carbon, nitrogen or phosphorus to some microorganisms [bib_ref] Removal of atrazine from river waters by indigenous microorganisms, Tappin [/bib_ref] , and alter nitrogen and/or phosphorus cycles [bib_ref] Biodegradation of carbaryl by a Micrococcus species, Doddamani [/bib_ref]. Thus, there is a need to understand the direct effect of pesticides on sediment nutrient dynamics, and how these changes can affect whole-ecosystem pools and fluxes of nutrients [bib_ref] Toward a metabolic theory of ecology, Brown [/bib_ref].
We measured the effects of atrazine, metolachlor, carbaryl and chlorothalonil on benthic microbial nutrient dynamics by quantifying net assimilation and remineralization rates of ammonium, nitrate and phosphate in laboratory mesocosms. We hypothesized that pesticides with a broad mode of action and higher affinity to organic matter such as chlorothalonil (disruption of cellular respiration, log K ow : 2.9) and metolachlor (inhibitor of mitosis and cellular division, log K ow : 3.4) would decrease microbial nitrate and phosphate uptake rates. In contrast, pesticides with more specific modes of action and higher affinity to the aqueous phase (atrazine: blocks photosynthesis, log K ow : 2.7 and carbaryl: inhibitor of synaptic activity, log K ow : 2.4) were predicted to have no effect on nutrient cycling. Further, natural resources managers and stakeholders would be able to make general predictions of agricultural pesticides effects on the microbial community based on the mode of action and water/solid affinities of these pesticides.
# Materials and methods
## Experimental mesocosms
Sediment and water collection was conducted in May 2012 at Ball State University field station property of Jakes Creek -Cooper farm/Skinner field and approval for these experiments was received following the appropriate procedures. This sample collection did not involve endangered or protected species. Jakes Creek is a 3rd order agriculturallyinfluenced stream with adjacent row crops (i.e. corn and soybean) in Muncie, Indiana within the Upper White River Watershed (UWRW). During the sampling time, Jakes Creek water temperature was 17uC, pH 7.92, depth 5 cm (at sampling location), and discharge was 37 L/s. -While stream samples were not collected for pesticides analysis at the time of this experiment; stream water and sediment samples were collected one week prior to this study at the same site that shows atrazine and metolachlor concentrations were below detection limits. Filtered water samples (ten 1000 mL and one 200 mL) were collected from the stream thalweg using a 60 mm syringe and subsequently filtered (Whatmanß glass fiber filter; 0.7 mm nominal pore size) into acidwashed Nalgeneß bottles. The 200 mL Nalgene plastic bottle was used to determine initial concentrations of nitrate, ammonium, and phosphate. A composite sediment sample (,2000 cm 3 ) was randomly collected from the top 5 cm of the stream benthos and placed into three Nalgene plastic bottles. Sediment samples were transported on ice and subsequently combined and homogenized using a USGS no. 5 sieve in the laboratory. Homogenized sediment (20 cm 3 ) and 60 ml filtered stream water were placed into each of 160 laboratory mesocosms (Fisherbrand sterile urine cup, 120 ml).
Stock solutions were prepared for atrazine (Atrazine 4L, 42.2% purity, Loveland, CO, US), metolachlor (Me-too-lachlor II, 84.4% purity, Drexel Chemical Company, TN, US), carbaryl (Sevin XLR Plus, 44.1% purity, Bayer, NC, US), and chlorothalonil (Bravo, 54% purity, Syngenta, NC, US) to achieve final stock concentrations of 10,000 mg/L for atrazine and metolachlor, 5,000 mg/L for carbaryl, and 8,000 mg/L for chlorothalonil. Aliquots from each stock solution were added to mesocosms to [bib_ref] Occurrence of pesticides in shallow groundwater of the United States: Initial results..., Kolpin [/bib_ref] [bib_ref] Pesticides in the nation's streams and ground water, Gilliom [/bib_ref] Detection frequency and concentrations of atrazine, metolachlor, carbaryl, and chlorothalonil in U.S. freshwaters. Detection frequency was estimated throughout the U.S. across 50 basins (33 agricultural, 10 urban and 7 mixed); mean and maximum concentrations correspond to 83 agricultural streams. Annual mean detection frequencies for each compound at each site provide the proportion of water samples that have detectable levels of pesticides for a year period. *Chlorothalonil was detected at concentrations of 290 mg/L in run-off near golf courses. doi:10.1371/journal.pone.0109190.t001 reach ten target treatment concentrations for each pesticide with four replicates for each treatment.
Mean and peak atrazine (2 mg/L, 201 mg/L) and metolachlor (1 mg/L, 78 mg/L) are detected at concentrations ,2-3 orders of magnitude higher than carbaryl (0.01 mg/L, 5 mg/L) or chlorothalonil (0.07 mg/L, 0.3 mg/L) throughout the U.S. [bib_ref] Occurrence of pesticides in shallow groundwater of the United States: Initial results..., Kolpin [/bib_ref] [bib_ref] Pesticides in the nation's streams and ground water, Gilliom [/bib_ref]. Thus, the treatment concentrations used were selected to include these environmentally relevant concentrations and appropriately represent pesticide occurrence in streams. Treatment solutions used in this study ranged from 0 mg/L (control) to the maximum concentrations detected in U.S. freshwaters for atrazine (200 mg/ L), metolachlor (80 mg/L), carbaryl (4 mg/L) and chlorothalonil (0.5 mg/L) [fig_ref] Table 2: Toxicity and Octanol-water partition coefficient of atrazine, metolachlor, carbaryl and chlorothalonil to... [/fig_ref].
Water from each mesocosm was removed after 24 h using a 10 mL syringe, subsequently filtered as above and placed into vials (two analytical replicates ,5 ml) for analysis of nitrate and phosphate via ion chromatograph (DIONEX, ICS-3000). The colorimetric phenol-hypochlorite technique, [bib_ref] Determination of ammonia in seawater by the indophenol-blue method: Evaluation of the..., Aminot [/bib_ref] was used to quantify ammonium concentrations. Initial concentrations (background) of nitrate, ammonium, and phosphate were also analyzed following the analytical methods above. Sediment dry mass in each mesocosm was quantified using an analytical balance (OHAUS, Adventurer SL AS64).
# Data analysis
Nutrient uptake rates were calculated for phosphate, ammonium and nitrate as changes in concentration over time (24 h) per g of dry mass (sediment) in response to treatments as (22):
[formula] Nutrient uptake rate~( C f {C i ) : V T : gdm [/formula]
Where: Cf = Final concentration (mg/L); Ci = Initial concentration (mg/L); V = Volume (L) in the jarT = time (h); gdm = g dry mass (g). Negative nutrient uptake rates indicated net remineralization of nutrients and positive nutrient uptake rates indicated net assimilation of nutrients [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. Uptake rates were divided by the average of the control treatment for each pesticide (N = 16) to assess the effects of different treatments for a particular pesticides. Thus, in this study, a response ratio relative to controls was used as the response variable. Further, data were logtransformed to meet normality assumptions for statistical analyses. SigmaPlot ß 12.0 software was used for linear and nonlinear regression analyses of response to pesticide concentration. The Akaike Information criterion (AIC) was used to select the best fit model among the different polynomial candidate models. Further, to develop predictive models of microbial response to pesticides in agricultural waters mean and peak concentrations of metolachlor and chlorothalonil throughout U.S. freshwaters [fig_ref] Table 1: Detection frequency and concentrations of atrazine, metolachlor, carbaryl, and chlorothalonil in U [/fig_ref] were used to calculate nitrogen and phosphorus uptake rates.
# Results
Nitrate, phosphate, and ammonium uptake rates varied less than one order of magnitude across pesticide treatments [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. Phosphate uptake rates were three orders of magnitude greater than ammonium uptake rates across treatments, though both phosphate (mean = 180.56 mg gdm 21 h 21 ) and ammonium (mean = 0.34 mg gdm 21 h 21 ) uptake rates were net assimilative [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. In contrast, we observed net remineralization of nitrate (mean = 232.22 mg gdm 21 h 21 ) across pesticides [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. Overall, these nutrient dynamics are expected in nitrogensaturated agricultural ecosystems (i.e., microbial nitrate assimilation is saturated and PO 4 becomes a limiting nutrient).
## Nitrate dynamics
Control nitrate uptake rate was ,20x higher on average than nitrate uptake influenced by pesticides (i.e., increasing remineralization in presence of pesticides). Nitrate uptake rates in response to atrazine, metolachlor, and carbaryl treatments exposure ranged from net assimilation (consumption/removal) to remineralization (source/addition) in the water column of our mesocosms [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. In contrast, increasing concentrations of chlorothalonil yielded increasing nitrate remineralization (p = 0.02, r 2 = 0.83, . No other pesticides had significant effects on nitrate uptake rates (p.0.05).
## Phosphate dynamics
Phosphate uptake rates in response to pesticides exposure was ,100x higher than control phosphate uptake rates. Thus, phosphate assimilation increased in the presence of atrazine, metolachlor, carbaryl, and chlorothalonil relative to the control [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. Further, phosphate uptake was negatively related to increasing chlorothalonil concentration (p,0.001, r 2 = 0.87, . Similarly, the metolachlor effect on phosphate uptake rate (p = 0.005, r 2 = 0.91, followed a cubic relationship with increasing phosphate uptake at lower concentrations (0 to 10 mg/L), and decreasing rates at higher concentrations (10 to 80 mg/L). No other pesticides resulted in significant effects on phosphate uptake rates (p.0.05). Ammonium dynamics
Ammonium uptake rates in response to pesticides exposure was ,10x higher than control ammonium uptake rate. Thus ammonium assimilation increased in the presence of atrazine, metolachlor, carbaryl, and chlorothalonil relative to the control [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. The ammonium uptake rate varied in response to metolachlor treatments (p = 0.023, r 2 = 0.83, , increasing with lower concentrations of metolachlor (0 to 10 mg/L), followed by a decline at higher concentrations (10 to 80 mg/L). No other pesticides had significant effects on ammonium uptake rates (p. 0.05). Predicting nutrient response to pesticides Changes in stream ecosystem nitrogen and phosphorus uptake rates were modeled across metolachlor and chlorothalonil concentrations detected throughout the U.S. . These changes were generated with polynomial regressions from this experiment (p,0.05).
The predicted ammonium uptake (0.52 NH 4 mg gdm 21 h 21 ) in response to mean metolachlor concentrations was ,20% higher than ammonium uptake (0.42 NH 4 mg gdm 21 h 21 ) at peak metolachlor concentrations. At mean concentrations of metolachlor, the predicted phosphate uptake (212.2 PO 4 mg gdm 21 h 21 ) was ,40% lower than phosphate uptake (372.41 PO 4 mg gdm 21 h 21 ) at peak metolachlor concentrations. At mean and peak chlorothalonil concentrations, there is net remineralization of nitrate (271.99 NO 3 mg gdm 21 h 21 and 241.87 NO 3 mg gdm 21 h 21 , respectively). Further, there was net assimilation of phosphate in response to mean and peak concentrations of chlorothalonil (207.88 and 77.81 PO 4 mg gdm 21 h 21 , respectively).
Overall, there was decreased ammonium assimilation and remineralization of nitrate in the presence of peak concentrations of metolachlor and chlorothalonil . Phosphate assimilation increased at peak concentrations of metolachlor and decreased at peak concentrations of chlorothalonil . Peak chlorothalonil is predicted to have the greatest effect on phosphate and nitrate uptake rates, by decreasing net assimilation and remineralization, respectively, over 50% relative to mean concentrations of these pesticides.
# Discussion
Research focuses primarily on the impacts of agricultural pesticides on fish and invertebrates [bib_ref] Pesticide toxicity index for freshwater aquatic organisms, Munn [/bib_ref] , [bib_ref] Analyzing effects of pesticides on invertebrate communities in streams, Liess [/bib_ref] ; however, little is known about effects on microbial communities. Our ecotoxicological research showed that benthic microbes' nutrient uptake rate response is likely a function of pesticide chemical characteristics, and how these changes can affect nutrient dynamics, due to nutrient availability.
Variation in nutrient uptake rates is likely a result of differences in baseline nutrient concentrations and the biotic community. For example, control ammonium uptake rates in this study (0.026 mg NH 4 gdm 21 h 21 ) were ,3x lower than control ammonium uptake rates in Bunch and Bernot [bib_ref] Nitrate and ammonium uptake by natural stream sediment microbial communities in response..., Bunch [/bib_ref]. These results suggest dissimilar nutrient demand due to unique biotic communities and nutrient conditions across sampling sites and temporal variability.
In our study, mean ammonium uptake rate was ,1.5x higher than ammonium uptake under enriched nitrate conditions and ,2.5x lower under enriched ammonium conditions reported in Bunch and Bernot [bib_ref] Nitrate and ammonium uptake by natural stream sediment microbial communities in response..., Bunch [/bib_ref]. Thus, ammonium uptake is higher in the presence of pesticides than under enriched nitrate conditions and lower under enriched ammonium conditions, due to microbial adaptations to nutrient availability and metabolic responses in presence of pesticides. Our nitrate uptake results were similar to yields with net remineralization under enriched ammonium conditions reported in Bunch and Bernot [bib_ref] Nitrate and ammonium uptake by natural stream sediment microbial communities in response..., Bunch [/bib_ref] , though remineralization was 10 times lower than rates measured in response to pesticides in our study.
The effects of metolachlor and chlorothalonil on nutrient uptake rates suggest a unique biotic community at this site, represented by mostly heterotrophic benthic microbes. Our results suggest a toxic effect of metolachlor and chlorothalonil on the benthic microbial community that is reflected by the increased nitrate remineralization, and reduced assimilation of ammonium and phosphate. This increase in remineralization rates could be an outcome of cellular lysis or a stress mechanism [bib_ref] Gross and net nitrogen uptake and DON release in the euphotic zone..., Bronk [/bib_ref]. Further, a biotic community characterized by autotrophs [bib_ref] The influence of ammonium, nitrate, and dissolved oxygen concentrations on uptake, nitrification,..., Kemp [/bib_ref] had ammonium and nitrate uptake rates three orders of magnitude higher than benthic uptake in our study; possibly, under these study conditions, primary producers have a higher assimilation rate than heterotrophic benthic microbes in response to the available forms of nitrogen [bib_ref] The influence of ammonium, nitrate, and dissolved oxygen concentrations on uptake, nitrification,..., Kemp [/bib_ref]. Further, control phosphate uptake rates in our study were ,5x lower than rates previously measured with phosphorus enrichment (0.1-2 mg/L; 34). Thus, phosphate is likely a limiting nutrient in our system. However, in presence of pesticides, phosphate uptake was two orders of magnitude higher than the rates measured under phosphorus limiting conditions [bib_ref] Factors controlling phosphorus limitation in stream sediments, Klotz [/bib_ref] possibly, in addition to available phosphorus in the water column, microbes were potentially degrading pesticides as a source of phosphorus [bib_ref] Phosphorus-containing pesticide breakdown products: quantitative utilization as phosphorus sources by bacteria, Cook [/bib_ref].
Nutrient dynamics in the presence of pesticides are dependent on the physicochemical characteristics (e.g. sorption kinetics, modes of action) of each pesticide. Sorption kinetics of pesticides and their corresponding index (Octanol-water partition coefficient, K ow ) determine the affinity of organic contaminants to either the water column or sediments [bib_ref] Pesticide soil sorption parameters: theory, measurement, uses, limitations and reliability, Wauchope [/bib_ref]. Atrazine and carbaryl have a higher affinity to the aqueous phase (Log K ow : 2.7 and 2.4, respectively), relative to metolachlor and chlorothalonil (Log K ow : 3.4 and 2.9, respectively). Thus, atrazine and carbaryl are likely more prevalent in the water column and less available to the sediment microbial communities, with minimal effect on benthic microbial activity. In contrast, metolachlor and chlorothalonil have a higher affinity to solids and higher prevalence in sediment, potentially affecting benthic microbial activity. Our interpretations of pesticide availability based on their water/solid affinities supports our results of no effects for atrazine and carbaryl (Log K ow ,2.7, p.0.05) within the tested range, decreasing nitrate remineralization and decreasing phosphate uptake in response to chlorothalonil, and decreasing ammonium and phosphate uptake in response to metolachlor (Log K ow .2.9, p,0.05).
Nutrient dynamics are also affected by the pesticide mode of action. Atrazine and carbaryl have specific modes of action; atrazine blocks photosynthesis, and carbaryl inhibits the activity of acetylcholinesterase, an enzyme of insects, fish, mammals [bib_ref] Natural factors to consider when using acetylcholinesterase activity as neurotoxicity biomarker in..., Durieux [/bib_ref]. Thus, the specificity of atrazine and carbaryl, and a sediment microbial community dominated by heterotrophs with no synaptic activity [bib_ref] pmoA-based analysis of methanotrophs in a littoral lake sediment reveals a diverse..., Pester [/bib_ref] may explain the lack of significant effects of these pesticides on benthic microbes within the tested range [fig_ref] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure [/fig_ref]. In contrast, metolachlor and chlorothalonil are broad spectrum pesticides . In our study, ammonium and phosphate uptake rates decreased with increasing metolachlor concentrations. These decreasing ammonium and phosphate uptake rates could be due to metolachlor inhibition of mitosis and cell division. Similarly, phosphate and nitrate uptake rates were affected by chlorothalonil; possibly this fungicide affects benthic microbes by disrupting cellular respiration. Thus, pesticides with non-specific modes of action (e.g. metolachlor, chlorothalonil) are more likely to have a significant effect on nutrient dynamics of sediment microbes, consistent with our study results.
Metolachlor and chlorothalonil not only affect nutrient dynamics of the sediment microbial community; they can also affect other processes. In our study, at increasing concentrations of chlorothalonil, there is a decrease in phosphate uptake, thus there is more phosphate available for organismal consumption. In agricultural streams, where phosphorus is the limiting nutrient [bib_ref] Phosphorus: a rate limiting nutrient in surface waters, Correll [/bib_ref] , increasing availability of phosphate can lead to algal blooms, eutrophication, hypoxia, loss of biodiversity (e.g. fish kills), and loss of aesthetic value of these habitats [bib_ref] Controlling eutrophication: nitrogen and phosphorus, Conley [/bib_ref]. In contrast, in the presence of peak concentrations of metolachlor, there is an increase in phosphate assimilation, which could mitigate the excess phosphate in streams [bib_ref] The role of microorganisms in mobilization and fixation of phosphorus in sediments, Gä Chter [/bib_ref]. Similarly, there is an increase of ammonium availability due to inhibited uptake rates in response to metolachlor, which in turn could potentially increase biological activity of pesticide resistant microbes [bib_ref] Characterisation of new strains of atrazine-degrading Nocardioides sp. isolated from Japanese riverbed..., Satsuma [/bib_ref] , [bib_ref] Removal of atrazine from river waters by indigenous microorganisms, Tappin [/bib_ref] , [bib_ref] Biodegradation of carbaryl by a Micrococcus species, Doddamani [/bib_ref] , [bib_ref] Phosphorus-containing pesticide breakdown products: quantitative utilization as phosphorus sources by bacteria, Cook [/bib_ref]. Nitrate remineralization also decreases in response to chlorothalonil, reducing nitrate availability, and potentially further mitigating excess nitrogen in these habitats [bib_ref] Microbial immobilization and the retention of anthropogenic nitrate in a northern hardwood..., Zogg [/bib_ref].
Mesocosms have a critical role in understanding the mechanisms driving ecological processes, [bib_ref] Moving on up: can results from simple aquatic mesocosm experiments be applied..., Spivak [/bib_ref] and provide a bridge between smaller, better controlled experiments and the larger freshwater ecosystems [bib_ref] Mesocosm Experiments as a Tool for Ecological Climate-Change Research, Stewart [/bib_ref]. For example, [bib_ref] Moving on up: can results from simple aquatic mesocosm experiments be applied..., Spivak [/bib_ref] revealed that the effects of nutrients on primary producers are similar in artificial habitats across five orders of magnitude in size. Also, [bib_ref] Mesocosm Experiments as a Tool for Ecological Climate-Change Research, Stewart [/bib_ref] mentioned that mesocosms help disentangle direct from indirect effects over scale. Thus, our conclusions try to bridge what we observed in the laboratory level and what could potentially occur at an agricultural influenced stream level.
Pesticide occurrence and concentrations in streams are dictated primarily by land-use. Streams receiving run-off from agricultural landscapes frequently have the highest concentrations of pesticides, compared to forested, mixed-use and urban lands. These high concentrations occur as pulses that are coupled by the seasonality of agricultural practices. Peak concentrations of metolachlor and chlorothalonil are detected from April to August. Thus, the effects of these pesticides on nutrient dynamics are highest during this critical time. At peak metolachlor and chlorothalonil concentrations there is a decrease of predicted ammonium assimilation and nitrate remineralization. Further, at peak concentrations of metolachlor there is increased phosphate assimilation and at peak concentrations of chlorothalonil there is decreased phosphate assimilation .
Our findings demonstrate individual effects of these pesticides on sediment nutrient dynamics that are likely driven by a pesticides' mode of action and water/sediment affinities. More studies are required to understand the net effect on ecosystems and to address their synergistic or antagonistic effects as mixtures. The regression equations we generated can complement models of nitrogen and phosphorus availability in streams to predict the potential changes in nutrient dynamics in response to increasing presence of pesticides in lotic ecosystems.
[table] Table 1: Detection frequency and concentrations of atrazine, metolachlor, carbaryl, and chlorothalonil in U.S. freshwaters. [/table]
[table] Table 2: Toxicity and Octanol-water partition coefficient of atrazine, metolachlor, carbaryl and chlorothalonil to daphnids, green algae, and humans. [/table]
[table] Table 3: Uptake rates for nitrate, phosphate, and ammonium in response to pesticides exposure. [/table]
|
Orientation of immobilized antigens on common surfaces by a simple computational model: Exposition of SARS-CoV-2 Spike protein RBD epitopes
A B S T R A C TThe possibility of immobilizing a protein with antigenic properties on a solid support offers significant possibilities in the development of immunosensors and vaccine formulations. For both applications, the orientation of the antigen should ensure ready accessibility of the antibodies to the epitope. However, an experimental assessment of the orientational preferences necessarily proceeds through the preparation/isolation of the antigen, the immobilization on different surfaces and one or more biophysical characterization steps. To predict a priori whether favorable orientations can be achieved or not would allow one to select the most promising experimental routes, partly mitigating the time cost towards the final product. In this manuscript, we apply a simple computational model, based on united-residue modelling, to the prediction of the orientation of the receptor binding domain of the SARS-CoV-2 spike protein on surfaces commonly used in lateral-flow devices. These calculations can account for the experimental observation that direct immobilization on gold gives sufficient exposure of the epitope to obtain a response in immunochemical assays.
# Introduction
The activity and reactivity of an immobilized protein strongly depend on its orientation with respect to the surface of the support in -or on -which it is immobilized. This holds true for enzymes, as well as for antibodies and antigens. Therefore, the possibility to control and manipulate the exposition of the relevant residues and protein surfaces plays an important role in the rational design of devices based on immobilized proteins. Among these devices, immunosensors represent an expanding space for research and market opportunities.
While the path to reach a technologically relevant product must rely upon a strong experimental characterization [bib_ref] Advanced characterization of immobilized enzymes as heterogeneous biocatalysts, Bolivar [/bib_ref] [bib_ref] On the relationship between structure and catalytic effectiveness in solid surface-immobilized enzymes:..., Bolivar [/bib_ref] , possibly relying upon atomic-level methodologies [bib_ref] Trapping RNase A on MCM41 pores: effects on structure stability, product inhibition..., Matlahov [/bib_ref] [bib_ref] Proton-detected solid-state NMR spectroscopy of bone with ultrafast magic angle spinning, Mroue [/bib_ref] [bib_ref] High-resolution structural characterization of a heterogeneous biocatalyst using solid-state NMR, Varghese [/bib_ref] [bib_ref] Conformations and intermolecular interactions in cellulose/silk fibroin blend films: a solid-state NMR..., Tian [/bib_ref] [bib_ref] Ubiquitin immobilized on mesoporous MCM41 silica surfaces -Analysis by solid-state NMR with..., Adiram-Filiba [/bib_ref] [bib_ref] Engineering lasparaginase for spontaneous formation of calcium phosphate bioinspired microreactors, Louka [/bib_ref] , the preparation/isolation of the protein of interest, its immobilization, and the characterization of the resulting composite are complex and time-consuming, therefore it is also true that guidelines for achieving optimal orientations could improve the efficiency of the R&D connected to protein immobilization [bib_ref] Computer-aided design of bromelain and papain covalent immobilization, Cutiño-Avila [/bib_ref] [bib_ref] Rational design strategy as a novel immobilization methodology applied to lipases and..., Monte-Martínez [/bib_ref]. However, simplified simulation models that would allow for a rapid prediction of the most plausible orientations are not particularly common. In this manuscript we apply a very simple method based on a united-residue modelling of protein-surface interactions, to specifically address the problem of determining the orientation of the SARS-CoV-2 Spike protein Receptor Binding Domain (RBD) on a few prototypical surfaces for biomedical use. United residue modelling of protein-surface interactions is a rather effective model to screen the poses of protein molecules with respect to surfaces [bib_ref] Orientation of adsorbed antibodies on charged surfaces by computer simulation based on..., Zhou [/bib_ref] [bib_ref] Parallel tempering Monte Carlo simulations of lysozyme orientation on charged surfaces, Xie [/bib_ref]. The method we apply is based on the works by Jiang, Zhou and del Monte-Martinez [bib_ref] Computer-aided design of bromelain and papain covalent immobilization, Cutiño-Avila [/bib_ref] [bib_ref] Orientation of adsorbed antibodies on charged surfaces by computer simulation based on..., Zhou [/bib_ref] [bib_ref] Parallel tempering Monte Carlo simulations of lysozyme orientation on charged surfaces, Xie [/bib_ref] [bib_ref] Molecular simulation studies of the orientation and conformation of cytochrome c adsorbed..., Zhou [/bib_ref] [bib_ref] Multiscale simulations of protein G B1 adsorbed on charged self-assembled monolayers, Liu [/bib_ref] [bib_ref] Lipase adsorption on different nanomaterials: a multiscale simulation study, Zhao [/bib_ref] [bib_ref] Ribonuclease A adsorption onto charged self-assembled monolayers: a multiscale simulation study, Liu [/bib_ref] , and encompasses van der Waals and electrostatic interactions, as well as covalent immobilization.
The choice of the target protein is motivated by the recent emergence of a new infectious disease (COVID-19) caused by a coronavirus (SARS-CoV-2). This infectious disease has spread significantly throughout the world, counting 13.841.890 infected people and a death toll of 590.845 as of July 2020 [bib_ref] An interactive web-based dashboard to track COVID-19 in real time, Dong [/bib_ref]. Models suggests that it will remain circulating and active for several months [bib_ref] Analysis and forecast of COVID-19 spreading in China, Fanelli [/bib_ref] [bib_ref] COVID-19: the unreasonable effectiveness of simple models, Carletti [/bib_ref] , and there is a marked possibility that reinfection is possible [bib_ref] The dynamics of humoral immune responses following SARS-CoV-2 infection and the potential..., Kellam [/bib_ref] , thus increasing the time of the circulation of the virus. This pandemic outbreak has had a major impact on world economics, with a very long outlook. A capillary control of the diffusion of the infection has proven crucial, and serological tests are expected to have a key role in mass screening [bib_ref] The important role of serology for COVID-19 control, Winter [/bib_ref] [bib_ref] Poolkeh Finds the Optimal Pooling Strategy for a Population-wide COVID-19 Testing (Israel,..., Eliaz [/bib_ref].
# Methods
The structures of the proteins were downloaded from the protein databank (PDB) [bib_ref] The Protein Data Bank, Berman [/bib_ref] , the pKa values of reactive groups were calculated using PROPKA [bib_ref] Very fast empirical prediction and rationalization of protein pKa values, Li [/bib_ref] [bib_ref] Improved treatment of ligands and coupling effects in empirical calculation and rationalization..., Søndergaard [/bib_ref] , and the interfaces were calculated using the PDBe PISA server [bib_ref] Inference of macromolecular assemblies from crystalline state, Krissinel [/bib_ref]. The non-bonded interaction of a residue of type i is represented with a Lennard-Jones (LJ) potential: [bib_ref] Orientation of adsorbed antibodies on charged surfaces by computer simulation based on..., Zhou [/bib_ref] [bib_ref] Parallel tempering Monte Carlo simulations of lysozyme orientation on charged surfaces, Xie [/bib_ref].
[formula] = + + U r r r ( ) 4 i i i i i 12 6 [/formula]
where r is the nearest distance between the residue and the surface, ε i is the energy at the minimum position, σ i is the equivalent van der Waals radius of each residue and δ i is a size parameter taken from the literature (see tables S2-S7, parameters are taken from [bib_ref] Molecular simulation studies of the orientation and conformation of cytochrome c adsorbed..., Zhou [/bib_ref] [bib_ref] Ribonuclease A adsorption onto charged self-assembled monolayers: a multiscale simulation study, Liu [/bib_ref] [bib_ref] Residue-residue potentials with a favorable contact pair term and an unfavorable high..., Miyazawa [/bib_ref] [bib_ref] Coarse-grained models for simulations of multiprotein complexes: application to ubiquitin binding, Kim [/bib_ref] [bib_ref] Generic coarse-grained model for protein folding and aggregation, Bereau [/bib_ref] [bib_ref] Low-resolution models for the interaction dynamics of coated gold nanoparticles with β2-microglobulin, Brancolini [/bib_ref] [bib_ref] Understanding the curvature effect of silica nanoparticles on lysozyme adsorption orientation and..., Yu [/bib_ref] , as indicated in the table captions).
The electrostatic interaction is represented through the Gouy-Chapman potential [bib_ref] Orientation of adsorbed antibodies on charged surfaces by computer simulation based on..., Zhou [/bib_ref] [bib_ref] Parallel tempering Monte Carlo simulations of lysozyme orientation on charged surfaces, Xie [/bib_ref] [bib_ref] Electrostatic interactions of a string-like particle with a charged plate, Tsao [/bib_ref].
[formula] = U r q e ( ) S i r r 0 [/formula]
where r is the nearest distance between the i-th residue with charge q i and the surface, σ S is the surface charge density, κ is the inverse Debye Length calculated from the ionic strength I as = I 0.304/ 1 , and the relative permittivity of the medium is assumed to be distancedependent (ε r = r) [bib_ref] Parallel tempering Monte Carlo simulations of lysozyme orientation on charged surfaces, Xie [/bib_ref] [bib_ref] Electrostatic interactions of a string-like particle with a charged plate, Tsao [/bib_ref]. A 1:1 buffer salt concentration of 0.15 mol dm −3 is assumed.
For silica, the surface charge density is estimated to be −0.3C m −2 [bib_ref] Understanding the curvature effect of silica nanoparticles on lysozyme adsorption orientation and..., Yu [/bib_ref].
For self-assembled charged monolayers (SAM), the charge density is set to +0.02C m −2 for the amino-capped monolayer (SAM-NH 2 ) and to −0.02C m −2 for the carboxyl-capped monolayer (SAM-CO 2 H) [bib_ref] Molecular simulation studies of the orientation and conformation of cytochrome c adsorbed..., Zhou [/bib_ref] [bib_ref] Ribonuclease A adsorption onto charged self-assembled monolayers: a multiscale simulation study, Liu [/bib_ref].
The formation of a covalent bond is treated with the following potential:
[formula] = U r if r elsewhere ( ) 0 B i [/formula]
where ε B is the bond energy and is set to 600 kJ mol −1 for imino bondsand 100 kJ mol −1 for gold-thiol bonds [bib_ref] Use of electroactive thiols to study the formation and exchange of alkanethiol..., Collard [/bib_ref] [bib_ref] The gold-sulfur interface at the nanoscale, Häkkinen [/bib_ref] [bib_ref] Quantifying thiol-gold interactions towards the efficient strength control, Xue [/bib_ref] , regardless of the starting oxidation state of the thiol [bib_ref] Thiols and disulfides on the Au(111) surface: the headgroup−gold interaction, Grönbeck [/bib_ref]. The desolvation energy is already accounted for in the vdW term.
LJ parameters for epoxide-glyoxyl functionalization is assumed to be equal to SAM-CO 2 H, whereas for gold the parameters have been adapted from reference [bib_ref] Low-resolution models for the interaction dynamics of coated gold nanoparticles with β2-microglobulin, Brancolini [/bib_ref].
The sampling of the relative protein-surface orientations is performed by rotating a plane around the center of mass of the protein. The plane is initially parallel to the z = 0 plane. Only two rotations are necessary, as all the rotations around the normal to the plane will yield the same energy. The first rotation by an angle α ∈ [0, π] is applied around the y-axis, followed by another rotation of an angle β ∈ [0, 2π] around the z-axis, and then a translation is applied to optimize the position, similarly to what is done in the popular PALES software [bib_ref] Prediction of sterically induced alignment in a dilute liquid crystalline phase: aid..., Zweckstetter [/bib_ref] [bib_ref] Prediction of charge-induced molecular alignment of biomolecules dissolved in dilute liquid-crystalline phases, Zweckstetter [/bib_ref] [bib_ref] NMR: prediction of molecular alignment from structure using the PALES software, Zweckstetter [/bib_ref]. The sampling of the α, β pairs is made uniform by using REPULSION angular sampling [bib_ref] REPULSION, A Novel Approach to Efficient Powder Averaging in Solid-State NMR, Bak [/bib_ref] [bib_ref] SIMPSON: a general simulation program for solid-state NMR spectroscopy, Bak [/bib_ref] [bib_ref] Computer-intensive simulation of solid-state NMR experiments using SIMPSON, Tošner [/bib_ref]. The distance of the plane to the protein is then set by minimizing the energy terms described above.
# Results and discussion
The most important feature of a composite thought for immunochemical applications is that the orientation of the antigen with respect to the surface must ensure the accessibility of the epitope to the antibodies, to guarantee the recognition. Therefore, we have selected the crystallographic structure of RBD in complex with a fragment (FAB) of the human antibody CR3022 (PDB ID: 6W41) [bib_ref] A Highly Conserved Cryptic Epitope in the Receptor-Binding Domains of SARS-CoV-2 and..., Yuan [/bib_ref] , and identified the interface residues relevant for the interaction [fig_ref] Figure 1: Biological assembly from the crystallographic structure 6 W41 [/fig_ref] and . In the analysis of the orientations, we assume that the full length antibody will have the same accessibility as the FAB because of the high flexibility of the linkers of the heavy chains (see [fig_ref] Figure 1: Biological assembly from the crystallographic structure 6 W41 [/fig_ref] [bib_ref] Refined structure of an intact IgG2a monoclonal antibody, Harris [/bib_ref] [bib_ref] Crystallographic structure of an intact IgG1 monoclonal antibody11Edited by, Harris [/bib_ref] [bib_ref] Crystal structure of a neutralizing human IgG against HIV-1: a template for..., Saphire [/bib_ref] [bib_ref] Structure of full-length human anti-PD1 therapeutic IgG4 antibody pembrolizumab, Scapin [/bib_ref]. It is also important to note that, while the spike protein is highly glycosylated at N-and O-positions [bib_ref] Deducing the N-and O-glycosylation profile of the spike protein of novel coronavirus..., Shajahan [/bib_ref] [bib_ref] Structure, function, and antigenicity of the SARS-CoV-2 spike glycoprotein, Walls [/bib_ref] , the structured part of the RBD which is recognized by the antibody only carries one glycation at position 343 [bib_ref] Deducing the N-and O-glycosylation profile of the spike protein of novel coronavirus..., Shajahan [/bib_ref] (pink in [fig_ref] Figure 1: Biological assembly from the crystallographic structure 6 W41 [/fig_ref] , and the glycation site faces away from the antibody binding site. On these grounds, we have not considered glycation (experimentally, this would be done expressing recombinant RBD in prokaryotic cells, whereas glycation would be obtained in human cells [bib_ref] Deducing the N-and O-glycosylation profile of the spike protein of novel coronavirus..., Shajahan [/bib_ref].
## Interaction with a hydrophobic surface
Hydrophobic adsorption occurs selectively on hydrophobic carriers at low ionic strength [bib_ref] Hydrolysis of edible oils by lipases immobilized on hydrophobic supports: effects of..., Al-Duri [/bib_ref]. It is a rather common immobilization protocol, because of its simplicity. The interaction is here represented only through a simple Lennard-Jones (LJ) potential, the parameters of which have been defined according to the hydrophobicity index (table S2) [bib_ref] Residue-residue potentials with a favorable contact pair term and an unfavorable high..., Miyazawa [/bib_ref] [bib_ref] Coarse-grained models for simulations of multiprotein complexes: application to ubiquitin binding, Kim [/bib_ref] [bib_ref] Generic coarse-grained model for protein folding and aggregation, Bereau [/bib_ref] [bib_ref] Low-resolution models for the interaction dynamics of coated gold nanoparticles with β2-microglobulin, Brancolini [/bib_ref]. The most probable orientation (5‰ relative population) is shown in [fig_ref] Figure 2: Antigen-FAB complex shown in superposition with the most probable positioning of a... [/fig_ref] , with the surface represented as a disk. In this, and in the following representations, the interaction is calculated for the antigen alone, and then the complex is shown for examining the interference of the surface with the binding. Of the 2000 considered orientations, 263 are within 10% of the probability of the orientation shown in [fig_ref] Figure 2: Antigen-FAB complex shown in superposition with the most probable positioning of a... [/fig_ref]. Most of those orientations involve contacts between the interface residues and the surface and are therefore expected to be poorly efficient for the recognition. This is not completely unexpected; as hydrophobic carriers mimic the interfaces formed by the naturally occurring interfaces of the proteins.
## Interaction with charged surfaces
Also this immobilization strategy is rather common because of its simplicity. It is slightly less general, because the outcome strongly depends on the nature of the protein and of the surface. The electrostatic interaction of the i-th residue with the uniformly charged surface with a given charge density is estimated by the Gouy-Chapman potential [bib_ref] Orientation of adsorbed antibodies on charged surfaces by computer simulation based on..., Zhou [/bib_ref] [bib_ref] Parallel tempering Monte Carlo simulations of lysozyme orientation on charged surfaces, Xie [/bib_ref] , which is added to a LJ term. The parameters defining each system are listed in tables S3-S6.
We have considered the following surfaces:
1) silica -a common chromatographic support with high negative surface charge; 2) positively charged self-assembled monolayer (SAM), with amino capping of the chains [bib_ref] Molecular simulation studies of the orientation and conformation of cytochrome c adsorbed..., Zhou [/bib_ref] [bib_ref] Ribonuclease A adsorption onto charged self-assembled monolayers: a multiscale simulation study, Liu [/bib_ref] ; 3) negatively charged SAM, with carboxylic capping of the chains [bib_ref] Molecular simulation studies of the orientation and conformation of cytochrome c adsorbed..., Zhou [/bib_ref] [bib_ref] Ribonuclease A adsorption onto charged self-assembled monolayers: a multiscale simulation study, Liu [/bib_ref].
Supports #2 and #3 imply the possibility of colorimetric detection through gold, vide infra.
The most probable orientations are shown in [fig_ref] Figure 3: Antigen-FAB complex shown in superposition with the most probable positioning with respect... [/fig_ref]. The relative populations of the orientations shown in [fig_ref] Figure 3: Antigen-FAB complex shown in superposition with the most probable positioning with respect... [/fig_ref] are 100% for silica and SAM-NH 2 . For SAM-CO 2 H, the orientations in [fig_ref] Figure 3: Antigen-FAB complex shown in superposition with the most probable positioning with respect... [/fig_ref] is populated for about 25%, and there are other 6 orientations out of 2000 that have relative population above 2%, all within a few degrees from the one with highest relative population, except one that is more tilted, yielding a larger accessibility, with a relative population around 4% [fig_ref] Figure 2: Antigen-FAB complex shown in superposition with the most probable positioning of a... [/fig_ref].
It is apparent that only negatively charged surfaces allow for the exposition of the epitope, and this is anyway relatively marginal.
These results suggest that it would be nontrivial to achieve a good orientation relying upon adsorption, either based on hydrophobic or on charge interactions. Therefore, we have considered directed approaches based on stronger interactions. In particular, we have considered epoxide-glyoxyl (directed at primary amine moieties) [bib_ref] Stabilization of enzymes by multipoint covalent immobilization on supports activated with glyoxyl..., López-Gallego [/bib_ref] and gold (thiols and disulfide bridges).
The glyoxyl-based approach is quite popular for multipoint orientation-selective immobilization of proteins on surfaces. It involves a two-step mechanism, in the first step, the primary amine groups of the protein are allowed to react with the aldehyde groups to form Schiff base bonds, in the second step the bonds are reduced with sodium borohydride. This kind of immobilization has been simulated in a similar way as described by del Monte-Martìnez et al. [bib_ref] Rational design strategy as a novel immobilization methodology applied to lipases and..., Monte-Martínez [/bib_ref] , assuming a working pH = 7.5, to maximize the reactivity of the N-terminus and at the same time limiting the reactivity of lysine residues (see .
The choice of gold is also extremely popular, because of two reasons: the strong plasmonic response of gold, which causes a purple coloring of the bioconjugate, and because of the relatively easy manipulation required. Current SARS-CoV2 serological tests are indeed based on gold conjugates [bib_ref] Development and clinical application of a rapid IgM-IgG combined antibody test for..., Li [/bib_ref]. The conjugation to the surface is simulated in the same way as the amine-glyoxyl reaction, assuming that all cysteines are equally reactive towards gold (disulfide bridges can interact with gold to a comparable extent as thiols) [bib_ref] Thiols and disulfides on the Au(111) surface: the headgroup−gold interaction, Grönbeck [/bib_ref]. The resulting orientation has 100% relative population. Colloidal gold has a net negative surface charge [bib_ref] Determination of the surface charge density of colloidal gold nanoparticles using second..., Kumal [/bib_ref] , but including the electrostatic term has no impact on the recovered orientation.
In the epoxide-glyoxyl strategy, the conjugation appears to be mostly directed at the N-terminus, 1 which is facing away from the recognition interface but is not topologically very remote. Therefore, the epitope will only be partially exposed, whereas for the gold conjugation, ample access to the epitope is possible in the most probable orientation [fig_ref] Figure 4: Antigen-FAB complex shown in superposition with the most probable positioning with respect... [/fig_ref].
Finally, a completely different strategy could be applied for conjugation to (e.g.) gold nanoparticles: the use of a avidin-biotin affinity system. Biotinylation can be achieved through amine-specific reagents [bib_ref] Preferential labeling of αamino N-terminal groups in peptides by biotin: application to..., Sélo [/bib_ref] , and improvement in the selectivity can be achieved with minimal engineering of the sequence [bib_ref] Selective N-terminal acylation of peptides and proteins with a Gly-His tag sequence, Martos-Maldonado [/bib_ref]. Given that there is a rather substantial difference in the calculated pKas for the different amine sites (see table S7), it can be expected that, for pH values lower than 7, all lysine residues will be protonated and thus less reactive with probability higher than 99%. The N-terminus is not facing the interaction site (see [fig_ref] Figure 1: Biological assembly from the crystallographic structure 6 W41 [/fig_ref]. Therefore, selective biotinylation at the N-terminus is expected to be possible. In this case, the accessibility of the epitope is warranted if the interaction between the antigen and streptavidin, if at all possible, is sufficiently weak.
To explore this possibility we have performed an initial-stage docking using ZDOCK [bib_ref] ZDOCK server: interactive docking prediction of protein-protein complexes and symmetric multimers, Pierce [/bib_ref] , and inspected the first two elements that had a significantly higher ZDOCK score [fig_ref] Figure 3: Antigen-FAB complex shown in superposition with the most probable positioning with respect... [/fig_ref]. The possible interaction between the RBD and streptavidin was investigated also using HADDOCK2.4 [bib_ref] The HADDOCK2.2 web server: user-friendly integrative modeling of biomolecular complexes, Van Zundert [/bib_ref]. The protein-protein interface residues were predicted with CPORT [bib_ref] CPORT: a consensus Interface predictor and its performance in prediction-driven docking with..., De Vries [/bib_ref] , and then used as "active" and "passive" residues in the HADDOCK calculation. About 10 lowly populated clusters with weak energy were obtained; the most significant three with the lowest HADDOCK-scores are reported in [fig_ref] Figure 3: Antigen-FAB complex shown in superposition with the most probable positioning with respect... [/fig_ref] and their energies in . Both dockings indicate that, should the interaction occur, it would occur in a position that does not interfere with the antigen-antibody recognition.
# Conclusions
In this work, we describe the use of united-residue modelling for the prediction of the orientation of the receptor binding domain of the spike protein of the novel coronavirus SARS-CoV-2, a protein of high immunological relevance at the most commonly used surfaces for the preparation of lateral-flow immunochemical devices. With this simple, yet very flexible approach, we find that immobilization on silica, or through glyoxyl reaction of amine residues, or on gold yield orientations compatible with antibody recognition, with gold granting the highest exposition. In this way, we can explain why random conjugation of the RBD to a gold surface yields responsive immunosensors, which are now routinely used. A more detailed experimental verification of the predictions of protein orientation at surfaces represents a significant challenge for the current biophysical methodologies [bib_ref] Advanced characterization of immobilized enzymes as heterogeneous biocatalysts, Bolivar [/bib_ref]. One can expect that cryo-electron transmission microscopy will be limited by the fact that, in most cases, the surface has higher electron density than that of the protein. Confocal laser scanning microscopy can be used to assess the positioning of the protein with respect to the support, and super-resolution microscopic techniques, such as total internal reflection fluorescence microscopy also allow for the detection of discrete molecular events (e.g., desorption, unfolding, lateral diffusion, …) [bib_ref] On the relationship between structure and catalytic effectiveness in solid surface-immobilized enzymes:..., Bolivar [/bib_ref] , but the orientation is still a high-hanging fruit by these methodologies. Conversely, the interaction between the protein and the interface can be probed at the atomic level through the application of solid-state NMR [bib_ref] Proton-detected solid-state NMR spectroscopy of bone with ultrafast magic angle spinning, Mroue [/bib_ref] [bib_ref] Ubiquitin immobilized on mesoporous MCM41 silica surfaces -Analysis by solid-state NMR with..., Adiram-Filiba [/bib_ref] [bib_ref] Engineering lasparaginase for spontaneous formation of calcium phosphate bioinspired microreactors, Louka [/bib_ref] [bib_ref] Atomic structural details of a protein grafted onto gold nanoparticles, Giuntini [/bib_ref] [bib_ref] 1H-detected solid-state NMR of proteins entrapped in bioinspired silica: a new tool..., Ravera [/bib_ref] [bib_ref] Probing the transmembrane structure and dynamics of microsomal NADPH-cytochrome P450 oxidoreductase by..., Huang [/bib_ref] , effort which is being started in our lab. Our results suggest that very simple modelling approaches can provide significant hints towards rationally orienting antigens in a way to maximize the exposition of epitopes, and therefore help in the initial moments of the design of conjugates for immunologic applications, when a rapid response to emergency is vital. This is also testified by the emergence of theoretical modelling of several molecular aspects of viral infection and inhibition mechanisms [bib_ref] Reckoning a fungal metabolite, Pyranonigrin a as a potential Main protease (Mpro)..., Rao [/bib_ref] [bib_ref] Identification of a novel dual-target scaffold for 3CLpro and RdRp proteins of..., Aouidate [/bib_ref] [bib_ref] Identification of potential binders of the main protease 3CLpro of the COVID-19..., Macchiagodena [/bib_ref] [bib_ref] Interaction of hydroxychloroquine with SARS-CoV2 functional proteins using all-atoms nonequilibrium alchemical simulations, Procacci [/bib_ref] [bib_ref] Delving deep into the structural aspects of a furin cleavage site inserted..., Li [/bib_ref]. Overall, the expected short-time impact of our work is to provide guidelines to avoid the experimental exploration of immobilization pathways that are less promising.
## Declaration of competing interest
The authors declare no competing financial interest.
[fig] Figure 1: Biological assembly from the crystallographic structure 6 W41. The SARS-CoV-2 receptor binding domain is shown in light green, with the interacting residues highlighted in red and the N-terminus highlighted in green. The glycation site 343 is highlighted in pink. The fragment of the human antibody CR3022 is shown in blue. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) [/fig]
[fig] Figure 2: Antigen-FAB complex shown in superposition with the most probable positioning of a hydrophobic surface. The surface is represented as a disk, aligned with the viewer. [/fig]
[fig] Figure 3: Antigen-FAB complex shown in superposition with the most probable positioning with respect to charged surfaces: (a) silica, (b) SAM-CO 2 H (negative) and (c) SAM-NH 2 (positive). [/fig]
[fig] Figure 4: Antigen-FAB complex shown in superposition with the most probable positioning with respect to covalently bound surfaces (a) epoxide-glyoxyl and (b) gold. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) [/fig]
|
The Interaction Between Non-Coding RNAs and Calcium Binding Proteins
Calcium binding proteins (CBP) are a group of proteins mediating the effects of calcium on cellular functions. These proteins can regulate calcium levels inside the cells and contribute in several cellular functions through transporting this ion across cell membranes or decoding related signals. Recent studies have shown that several noncoding RNAs interact with CBPs to affect their expression or activity. The interactions between these transcripts and CBPs have implications in the pathoetiology of human disorders, including both neoplastic and non-neoplastic conditions. In the current review, we describe the interactions between three classes of non-coding RNAs (long non-coding RNAs, circular RNAs, and microRNAs) and a number of CBPs, particularly CAB39, S100A1, S100A4, S100A7 and S100P. This kind of interaction has been verified in different pathological contexts such as drug-induced cardiotoxicity, osteoblasts cytotoxicity, acute lung injury, myocardial ischemia/reperfusion injury, proliferative diabetic retinopathy, glomerulonephritis, as well as a wide array of neoplastic conditions.
# Introduction
Calcium is an important second messenger in cells whose effects are largely dependent on a number of diverse proteins, being named as calcium binding proteins (CBP), accordingly. These proteins can bind this ion in their certain domains. CBPs can regulate calcium levels inside the cells and contribute in several cellular functions through transporting this ion across cell membranes or decoding related signals [bib_ref] Calcium Binding Proteins, Yañez [/bib_ref]. Based on the presence of the structural EF-hand domain, intracellular CBPs can be classified into two main classes, i.e., those containing this domain and those lacking this domain. Parvalbumin, calmodulin, S100 proteins and calcineurin are examples of the former class, while calreticulin, calsequestrin, annexins, protein kinase C (PKC) and sinaptotagmin are examples of the latter [bib_ref] Calcium Binding Proteins, Yañez [/bib_ref]. Extracellular CBP has six main classes, based on the presence of EF-hand, EGF-like, g-carboxyl glutamic acid (GLA)-rich, cadherin, and calcium-dependent (C)-type lectin-like domains or calcium binding pockets of family C G-proteincoupled receptors [bib_ref] Calcium Binding Proteins, Yañez [/bib_ref]. Extracellular CBPs are incessantly surrounded by a concentration of 10 −3 M calcium which contributes in the activation or stabilization of specific enzymes acting as protease, nuclease, or lipase. On the other hand, intracellular CBPs, which act as muscle contraction, respond to an upsurge in calcium concentrations from 10 −7 to 10 −6 M (2). Recent studies have shown that several non-coding RNAs interact with CBPs to affect their expression or activity. The interactions between these transcripts and CBPs have implications in the pathoetiology of human disorders, including both neoplastic and non-neoplastic conditions. In the current review, we describe the interactions between three classes of non-coding RNAs (long non-coding RNAs (lncRNAs), circular RNAs [circRNAs) and microRNAs (miRNAs)] and CBPs.
## Non-coding rnas and cab39
Calcium-binding protein 39 (CAB39) is functionally associated with the Serine/Threonine Kinase STK11 and STRAD [bib_ref] MO25alpha/beta Interact With STRADalpha/beta Enhancing Their Ability to Bind, Activate and Localize..., Boudeau [/bib_ref]. This protein also promotes the construction of STK11/STRAD complexes and induces catalytic activity of STK11 (3). This protein has been found to affect the process of doxorubicininduced cardiac injury. Experiments in an animal model of doxorubicin-induced cardiotoxicity have shown up-regulation of miR-451 levels. Suppression of miR-451 expression has reduced doxorubicin-associated whole-body wasting and cardiac atrophy, decreased heart damage, amended heart function, and enhanced contractile function of cardiomyocytes. Functionally, miR-451 suppression has led to enhancement of Cab39 levels and induced activity of AMPK signaling . Thus, Cab39 has been identified as the target of miR-451 through which this miRNA affects cardiac toxicity [bib_ref] Silencing Inhibited Doxorubicin Exposure-Induced Cardiotoxicity in Mice, Li [/bib_ref]. Another study to find the mechanism of osteoblast cytoprotection has reported miR-107 as a CAB39-targeting miRNA. Functional experiments in OB-6 human osteoblastic cells have shown direct binding of this miRNA with CAB39 mRNA. Both wild-type miR-107 mimics and pre-miR-107-containing lentiviruses could inhibit CAB39 expression in osteoblasts. On the other hand, miR-107 antagonism could increase CAB39 expression, leading to activation of AMPK cascade. Suppression of miR-107 has significantly decreased dexamethasone-induced apoptosis in FIGURE 1 | The connection between CAB39 and miRNAs, as well as their role in human diseases. Inhibition of miRNA has resulted in increased CAB39 levels and increased activity of AMPK pathway. Cab39 has therefore been found as a miRNAs target, and these miRNAs modulate cardiotoxicity, osteoblasts, cytotoxicity, acute lung damage, chemoresistance, senescence, and cancer development through this RNA. OB-6 cells and human osteoblasts. Moreover, antagomiR-107 could activate AMPK downstream Nrf2 cascade to suppress dexamethasone-associated oxidative injury [bib_ref] miR-107 Inhibition Upregulates CAB39 and Activates AMPK-Nrf2 Signaling to Protect Osteoblasts From..., Zhuang [/bib_ref]. The interaction between miRNAs and CAB39 has also been implicated in the pathogenesis of acute lung injury. The CAB39-interacting miR-31-5p has been shown to be up-regulated in mice lung tissues upon injection of lipopolysaccharide. miR-31-5p silencing has relieved, while miR-31-5p mimic has aggravated lipopolysaccharide-induced inflammatory responses, oxidative injury, and pulmonary injury in vivo and in vitro. Functionally, miR-31-5p silencing has induced protective impact of AMPKa. In fact, Cab39 has an essential role in activation of AMPKa and protective effects of miR-31-5p antagomir [bib_ref] MicroRNA-31-5p Exacerbates Lipopolysaccharide-Induced Acute Lung Injury via Inactivating Cab39/Ampka Pathway, Jiang [/bib_ref]. The interaction between miRNAs and CAB39 has an important role in the pathogenesis of cancers. For instance, miR-1265 can regulate cell proliferation and apoptosis in gastric cancer cells by targeting CAB39. miR-1265-mediated suppression of CAB39 interferes with oncogenic autophagy through modulation of AMPK/mTOR [bib_ref] MIR-1265 Regulates Cellular Proliferation and Apoptosis by Targeting Calcium Binding Protein 39..., Xu [/bib_ref]. On the other hand, miR-107-mediated suppression of CAB39 and subsequent activation of AMPK/mTOR signaling confers chemoresistance to colorectal cancer [bib_ref] MiR-107 Confers Chemoresistance to Colorectal Cancer by Targeting Calcium-Binding Protein 39, Liang [/bib_ref]. [fig_ref] TABLE 1 |: Shows the interaction between CBP [/fig_ref] summarizes the role of CAB39-interacting miRNAs in the pathogenesis of different disorders.
CircGSK3B (hsa_circ_0003763) is a circRNA that has indirect interaction with CAB39. This circRNA has been found to be upregulated in hepatocellular cancer tissues and cell lines. In addition, expression levels of circGSK3B have been correlated with tumor bulk and vascular invasion. Functional studies have indicated the role of circGSK3B in the enhancement of proliferation, migratory potential, and invasiveness of hepatocellular carcinoma. Mechanistically, circGSK3B sponges miR-1265 to up-regulate expression of CAB39 . This circRNA has a role in reprogramming of glutamine metabolism. Taken together, circGSK3B/miR-1265/CAB39 axis has a role in enhancing proliferation, migration, invasion of this kind of cancer [bib_ref] Circular RNA Circgsk3b Promotes Cell Proliferation, Migration, and Invasion by Sponging miR-1265..., Li [/bib_ref]. HOTAIR is an example of lncRNAs that activates AMPKa via EZH2/miR-451/CAB39 axis regulation. CAB39 is involved in regulation of oxidative stress and cardiac myocyte apoptosis during ischemia/reperfusion injury [bib_ref] The Long Noncoding RNA Hotair Regulates Oxidative Stress and Cardiac Myocyte Apoptosis..., Meng [/bib_ref]. [fig_ref] TABLE 2 |: CAB39-interacting lncRNAs/circRNAs [/fig_ref] summarizes CAB39-interacting lncRNAs/circRNAs.
## Non-coding rnas and s100a4
S100A4 is a member of the S100 CBP family, which is produced by tumor cells as well as stromal cells. S100 proteins are localized in the cytoplasm and/or nucleus of several kinds of cells and participate in the regulation of cell cycle transition and differentiation. The genes coding members of S100 family are clustered on chromosome 1q21 [bib_ref] Six S100 Genes Are Clustered on Human Chromosome 1q21: Identification of Two..., Engelkamp [/bib_ref]. This CBP has been shown to support tumorigenesis through stimulation of angiogenesis. A number of miRNAs have been shown to inhibit the expression of S100A4. For instance, miR-187-3p by targeting S100A4 could inhibit the metastasis and epithelial-mesenchymal transition (EMT) of hepatocellular carcinoma [bib_ref] Deregulated miR-296/S100A4 Axis Promotes Tumor Invasion by Inducing Epithelial-Mesenchymal Transition in Human..., Dou [/bib_ref]. Expression of S100A4 has been shown to be increased in ovarian cancer in association with clinical stage of these patients. Down-regulation of this CBP Beclin-1, LC3, p62, AMPK/mTOR miR-1265 by targeting CAB39 could regulate and apoptosis in GC and impair autophagy.
-miR-34a-3p -dental pulp stem cells (DPSCs) p53, p21, p16, AMPK/ mTOR Metformin-induced miR-34a-3p downregulation by targeting CAB39 via the AMPK/ mTOR pathway could alleviate senescence in human DPSCs. [bib_ref] Metformin-Induced MicroRNA-34a-3p Downregulation Alleviates Senescence in Human Dental Pulp Stem Cells by..., Zhang [/bib_ref] has reduced the mobility of ovarian cancer cells and their metastatic ability, while up-regulation of S100A4 has increased the invasive aptitude of these cells. miR-296 has been identified as an important upstream regulator of this CBP . Dysregulation of miR-296/S100A4 axis could facilitate EMT (15). Another study in bladder cancer has revealed that miR-149-3p could inhibit proliferation, migration, and invasion of malignant cells through targeting S100A4 [bib_ref] Expression of miR-149-3p Inhibits Proliferation, Migration, and Invasion of Bladder Cancer by..., Yang [/bib_ref]. In colorectal cancer cells, miR-325-3p/S100A4 (17), miR-520c/S100A4 (18) and miR-296/S100A4 [bib_ref] Long non-Coding RNA PTENP1 Functions as a ceRNA to Modulate PTEN Level..., Zhang [/bib_ref] have been identified as molecular axes that affect carcinogenesis. [fig_ref] TABLE 3 |: S100A4-interacting miRNAs [/fig_ref] shows S100A4-interacting miRNAs.
A number of lncRNAs have been found to affect miRNA/S100A4 axes. These lncRNAs mainly act as molecular sponges for S100A4-interacting miRNAs, thus releasing S100A4 from inhibitory effects of these miRNAs. For instance, HOXA-AS2 through regulating miR-520c-3p/S100A4 (23) and miR-520c-3p/S100A4 (24) axes could affect pathogenesis of papillary thyroid cancer and acute myeloid leukemia, respectively. Moreover, Linc01833 via regulation of miR-519e-3p/S100A4 axis could enhance progression of lung cancer [bib_ref] The Long Noncoding RNA Linc01833 Enhances Lung Adenocarcinoma Progression via MiR-519e-3p/ S100A4..., Zhang [/bib_ref]. [fig_ref] TABLE 4 |: S100A4-interacting lncRNAs [/fig_ref] shows S100A4-interacting lncRNAs.
NON-CODING RNAs AND S100A7 S100A7 is another member of the S100 family of proteins which contains 2 EF-hand domains. S100A7 differs from the other FIGURE 2 | Interaction between CBPs (CAB39 and S100A4) and lncRNAs/circRNAs with their contribution in human disorders. members of this family in the absence of calcium binding capacity in one of its EF-hand domains, which is located at the N-terminus. S100A7 has been shown to regulate metastatic ability of ovarian cancer cells and chemoresistance phenotype through modulation of MAPK pathway . This CBP has been shown to be targeted by miR-330-5p [bib_ref] S100A7 Regulates Ovarian Cancer Cell Metastasis and Chemoresistance Through MAPK Signaling and..., Lin [/bib_ref]. Moreover, S100A7 has been found to be a direct target of miR-26b-5p. In fact, miR-26b-5p can suppress proliferation, migration and invasiveness of intrahepatic cholangiocarcinoma cells through deceasing expression of S100A7 [bib_ref] MicroRNA-26b-5p Regulates Cell Proliferation, Invasion and Metastasis in Human Intrahepatic Cholangiocarcinoma by..., Fan [/bib_ref]. [fig_ref] TABLE 5 |: S100A7-interacting miRNAs [/fig_ref] shows S100A7interacting miRNAs.
## Non-coding rnas and s100a1
Similar to the majority of S100 proteins, binding of S100A1 with calcium results in great conformational alterations which facilitate interaction of this CBP with several protein targets. Targets of this CBP are those participating in calcium-related signal transduction, neurotransmitter release, cytoskeletal and filament associated proteins, transcription factors, a number of different proteins with enzymatic functions, and other CBPs, particularly S100B, S100A4 and S100P [bib_ref] S100A1: Structure, Function, and Therapeutic Potential, Wright [/bib_ref]. Suppression of S100A1 expression has been suggested as a therapeutic modality for treatment of various disorders such as neurological disorders, diabetes mellitus, heart failure, and numerous kinds of malignancies [bib_ref] S100A1: Structure, Function, and Therapeutic Potential, Wright [/bib_ref]. FOXD2-AS1 is the only lncRNA whose interactions with S100A1 have been verified. This lncRNA through modulation of S100A1/Hippo (29) and miR-363-5p/S100A1 pathways [bib_ref] Long Non-Coding RNA FOXD2-AS1 Aggravates Nasopharyngeal Carcinoma Carcinogenesis by Modulating miR-363-5p/S100A1 Pathway, Chen [/bib_ref] can participate in the pathogenesis of breast cancer and nasopharyngeal carcinoma, respectively [fig_ref] TABLE 6 |: S100A1-interacting lncRNAs [/fig_ref].
## Non-coding rnas and s100p
S100P is another member of S100 CBPs that mediate calciumdependent signal transduction [bib_ref] S100: A Multigenic Family of Calcium-Modulated Proteins of the EF-Hand Type With..., Donato [/bib_ref]. S100P has been primarily isolated from the placenta [bib_ref] S100P, a Novel Ca2+-Binding Protein From Human Placenta: cDNA Cloning, Recombinant Protein..., Becker [/bib_ref]. From an evolutionary point of view, S100P is regarded as a novel gene, existing only in the vertebrate genomes [bib_ref] The Life and Works of S100P -From Conception to Cancer, Prica [/bib_ref]. As S100P is expressed in the uterus during the rhythmic hormonal changes, it might be associated with embryonic implantation/development [bib_ref] The Life and Works of S100P -From Conception to Cancer, Prica [/bib_ref]. Yet, the role of S100P has been mostly investigated in the context of cancer [bib_ref] The Life and Works of S100P -From Conception to Cancer, Prica [/bib_ref]. Two independent studies have assessed association between S100P and non-coding RNAs in pancreatic cancer. First, miR-495 has been shown to suppress pancreatic carcinogenesis by targeting S100P [bib_ref] S100P Acts as a Target of miR-495 in Pancreatic Cancer Through Bioinformatics..., Jiang [/bib_ref]. Secondly, circ_0092314 has been shown to FIGURE 3 | The interaction of CBPs (S100A4, S100A7, S100A16, S100A9) with miRNAs. Linc01833 via the miR-519e-3p/S100A4 axis could enhance LUAD progression. miR-363-5p FOXD2-AS1, by modulating miR-363-5p/S100A1 pathway, could participate in NPC carcinogenesis. [bib_ref] Long Non-Coding RNA FOXD2-AS1 Aggravates Nasopharyngeal Carcinoma Carcinogenesis by Modulating miR-363-5p/S100A1 Pathway, Chen [/bib_ref] induce EMT in this type of cancer through sponging miR-671 and releasing S100P from its inhibitory effects [bib_ref] CircRNA Circ_0092314 Induces Epithelial-Mesenchymal Transition of Pancreatic Cancer Cells via Elevating the..., Shen [/bib_ref]. [fig_ref] TABLE 7 |: S100P-ineracting miRNAs and circRNAs [/fig_ref] shows S100P-ineracting miRNAs and circRNAs.
## Non-coding rnas and other calcium binding proteins
S100A8, S100A9, S100A10, S100A11, S100A14, S100A16, NECAB3 and SMOC2 are other CBPs whose interactions with non-coding RNAs have been verified in the context of human disorders . LINC00174 via targeting regulates miR-320/S100A10 axis could increase malignant phenotypes [bib_ref] LINC00174 Is an Oncogenic lncRNA of Hepatocellular Carcinoma and Regulates miR-320/ S100A10..., Zhao [/bib_ref]. SNHG8 is another lncRNA which serves as a sponge for miR-1270 to up-regulate expression of S100A11 and promote progression of ovarian cancer [bib_ref] lncRNA SNHG8 Promotes Ovarian Cancer Progression Through Serving as Sponge for miR-1270..., Xuan [/bib_ref]. In the context of lung cancer, CASC9 has been found to sponge miR-335-3p and induce expression of S100A14 [bib_ref] Upregulated lncRNA CASC9 Contributes to Progression of Non-Small Cell Lung Cancer Through..., Zhao [/bib_ref]. In addition, GNAS-AS1 serves as a sponge for miR-4319 to increase expression of NECAB3 and regulate macrophage polarization [bib_ref] GNAS-AS1/miR-4319/NECAB3 Axis Promotes Migration and Invasion of Non-Small Cell Lung Cancer Cells..., Li [/bib_ref].
# Discussion
The interactions between ncRNAs and CBPs have been assessed in different contexts. Most of studies have been conducted in the context of cancer, where CBPs affect malignant features through a variety of mechanisms, particularly induction of EMT. CAB39 is among the mostly assessed CBPs in this regard. Notably, the functional effect of CAB39-interacting miRNAs on the cells is largely mediated through modulation of activity of AMPK/ mTOR. S100A4 as another CBP has been shown to affect expression of EMT-markers such as E-cadherin, Vimentin, Ncadherin and Snail1. A number of miRNAs such as miR-187-3p, miR-296, miR-149-3p, miR-19a, miR-325-3p, miR-520c and miR-296 have been shown to affect carcinogenesis through modulation of expression of S100A4. Thus, S100A4-interacting non-coding RNAs are putative targets for design of novel therapeutic options against tumor metastasis and EMT. S100P and S100A16 are other CBPs whose interactions with noncoding RNAs are implicated in the process of EMT. In fact, miRNAs that affect expression of CBPs have been shown to bind with 3' UTR of mRNAs coding for CBPs. circRNAs and lncRNAs that affect expression of CBPs mainly act as molecular sponges for miRNAs. For instance, circGSK3B/ miR-1265/CAB39, circ_0092314/miR-671/S100P, HOXA-AS2/ miR-520c-3p/S100A4, HOXA-AS2/miR-520c-3p/S100A4, Linc01833/miR-519e-3p/S100A4, LINC00174/miR-320/ S100A10, SNHG8/miR-1270/S100A11, CASC9/miR-335-3p/ S100A14 and GNAS-AS1/miR-4319/NECAB3 are examples of these regulatory axes which are involved in the pathoetiology of human disorders, particularly cancers.
The regulatory effects of some miRNAs on their specific CBPs have been verified in different contexts. For instance, the inhibitory impact of miR-451 on CAB39 has been shown to be implicated in drug-associated cardiac toxicity as well as lung
Chronic Sepsis miR-21, miR-181b S100A9 C57BL/6N S100a9 knockout mice Gr1+CD11b+ STAT3, C/EBPb, IL-10 S100A9 by inducing both miR-181b and miR-21 could maintain myeloidderived suppressor cells in chronic sepsis. miR-4319, IL-10, Arg-1 GNAS-AS1/miR-4319/NECAB3 axis by altering macrophage polarization could promote migration and invasion of NSCLC cells. [bib_ref] GNAS-AS1/miR-4319/NECAB3 Axis Promotes Migration and Invasion of Non-Small Cell Lung Cancer Cells..., Li [/bib_ref] cancer. Similarly, CAB39 has been found as a target of miR-107 in both osteoblasts and colorectal cancer cells. Finally, S100A4 has been shown to be targeted by miR-296 in both ovarian and colorectal cancer cells. For other miRNAs, regulatory effects have been confirmed only in a single context. Taken together, several miRNAs, lncRNAs and circRNAs can regulate expressions of CBPs and participate in the etiology of human disorders via this route. Identification of this type of interactions has practical significance in design of disorders which are associated with abnormal calcium signal transduction. Research in this field is still in its infancy and the functional associations between non-coding RNAs and several members of CBP family need to be clarified.
# Author contributions
SG-F wrote the draft and revised it. MT and AB designed and supervised the study. HS, JM, BH, and HH collected the data and designed the figures and tables. All the authors read and approved the submitted version.
[fig] FOXD2-: AS1 BALB/c nude mice, Dataset MCF-10A, MCF-7, BT-549, MDA-MB-468, MDA-MB-453Cyclin-E1, CDK2, p21, MMP2/9, YAD, LATS1, MST1/2, Hippo AS1 via the S100A1/Hippo signaling pathway could regulate the tumorigenesis and progression of BCa. [/fig]
[table] TABLE 1 |: Shows the interaction between CBP (CAB39) with miRNAs. [/table]
[table] TABLE 2 |: CAB39-interacting lncRNAs/circRNAs. [/table]
[table] TABLE 3 |: S100A4-interacting miRNAs. [/table]
[table] TABLE 4 |: S100A4-interacting lncRNAs. [/table]
[table] TABLE 5 |: S100A7-interacting miRNAs. [/table]
[table] TABLE 6 |: S100A1-interacting lncRNAs. [/table]
[table] TABLE 7 |: S100P-ineracting miRNAs and circRNAs. AKT Circ_0092314 via elevating S100P expression by sponging miR-671 could induce EMT. [/table]
[table] TABLE 8 |: miRNAs interacting with other calcium binding proteins.HEC-1A, 293T, HEC-1A/ Pax -miR-24 by targeted silencing of the S100A8 gene could act as a tumorsuppressing gene and increase chemotherapy sensitivity of EC cells to paclitaxel. [/table]
[table] TABLE 9 |: lncRNAs interacting with other calcium binding-proteins. IOSE, A2780, HOSE 11-12, SKOV3, HO8910, OVCAR3 miR-1270 SNHG8 via serving as a sponge for miR-1270 to regulate S100A11 could promote OC progression.Upregulation of CASC9 via inhibiting miR-335-3p and activating S100A14 could contribute to the progression of NSCLC. [/table]
|
Nonalcoholic fatty liver disease and risk of incident hypertension: a systematic review and meta-analysis
Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's website, www.eurojgh.comThis is an open access article distributed under the Creative Commons Attribution License 4.0 (CCBY), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Several studies reported an association between nonalcoholic fatty liver disease (NAFLD) and the risk of incident hypertension. The objective of this systematic review and meta-analysis was to obtain a precise and reliable estimate of the nature and magnitude of this association. We systematically searched Ovid-MEDLINE up to March 2021 for observational studies in which NAFLD was diagnosed in adults using blood-based panels, imaging techniques or liver biopsy and with a follow-up ≥1 year. Measures of association from individual studies were meta-analyzed using random-effects models. Of the 1108 titles initially scrutinized, we included 11 cohort studies with data on 390 348 participants (52% male) and a mean follow-up of 5.7 years. In the overall analysis, NAFLD was associated with a moderately increased risk of incident hypertension (hazard ratio 1.66; 95% confidence interval (CI), 1.38-2.01; test for overall effect z = 5.266; P < 0.001). There was significant heterogeneity among the studies (P < 0.001). Sensitivity analyses showed that estimates were not affected by geographical location, duration of follow-up and adjustment for baseline blood pressure values. On the other hand, the magnitude of the association was lower in studies that adjusted for baseline adiposity compared with those that did not, explaining part of the observed heterogeneity. No significant publication bias was detected by funnel plot analysis and Egger's and Begg's tests. This large meta-analysis indicates that NAFLD is associated with a ~1.6-fold increased risk of developing hypertension. Further studies are needed to investigate the role of NAFLD severity in terms of inflammation and fibrosis on incident hypertension. Eur J Gastroenterol Hepatol 34: 365-371
# Introduction
Nonalcoholic fatty liver disease (NAFLD) is the most common form of chronic liver disease, affecting ~25% of the adult world population [bib_ref] Global epidemiology of nonalcoholic fatty liver disease-Meta-analytic assessment of prevalence, incidence, and..., Younossi [/bib_ref]. It is an umbrella term including patients with different degrees of histologic severity spanning from simple steatosis to lobular inflammation and hepatocyte ballooning (nonalcoholic steatohepatitis) to collagen deposition leading to liver fibrosis and possibly cirrhosis [bib_ref] Non-alcoholic fatty liver disease: a review of epidemiology, risk factors, diagnosis and..., Huang [/bib_ref] [bib_ref] The diagnosis and management of nonalcoholic fatty liver disease: practice guidance from..., Chalasani [/bib_ref]. It is increasingly recognized as a frequent cause of liver-related morbidity and mortality and its global prevalence is expected to further increase in the foreseeable future, given the widespread rise in obesity rates among adolescents and young adults [bib_ref] Prevalence of liver steatosis and fibrosis detected by transient elastography in adolescents..., Ciardullo [/bib_ref].
Being frequently associated with insulin resistance and ectopic fat deposition, its prevalence is even higher in patients with metabolic disorders, such as type 2 diabetes [bib_ref] High prevalence of advanced liver fibrosis assessed by transient elastography among U.S...., Ciardullo [/bib_ref] , and in patients displaying signs of the metabolic syndrome, including visceral obesity, dyslipidemia and essential hypertension [bib_ref] Nonalcoholic fatty liver disease: an emerging driver of hypertension, Zhao [/bib_ref] [bib_ref] Nonalcoholic fatty liver disease and advanced fibrosis in US adults across blood..., Ciardullo [/bib_ref] [bib_ref] Blood pressure, glycemic status and advanced liver fibrosis assessed by transient elastography..., Ciardullo [/bib_ref].
Accumulating evidence suggests that NAFLD is associated with an increased prevalence and incidence of hypertension [bib_ref] Non-alcoholic fatty liver disease and hypertension: coprevalent or correlated?, Oikonomou [/bib_ref] [bib_ref] Association between MRI-derived hepatic fat fraction and blood pressure in participants without..., Lorbeer [/bib_ref] , which still represents by far the most common disease that affects human beings and is considered the top contributor to the burden of disease worldwide [bib_ref] ESC/ESH Guidelines for the management of arterial hypertension: The Task Force for..., Williams [/bib_ref] [bib_ref] The global epidemiology of hypertension, Mills [/bib_ref].
To date, two previous meta-analyses examined the association between γ-glutamyl transpeptidase (γ-GT) levels and risk of incident hypertension [bib_ref] Gamma-glutamyltransferase and risk of hypertension: a systematic review and dose-response meta-analysis of..., Kunutsor [/bib_ref] [bib_ref] Gamma-glutamyltransferase level and risk of hypertension: a systematic review and meta-analysis, Liu [/bib_ref] , whereas no quantitative summary of the available evidence is present on studies using more accurate measures of liver fat content such as specific blood-based panels, imaging techniques or liver biopsy.
We have therefore undertaken a systematic review and meta-analysis of observational cohort studies of adults from different geographical locations examining the association between NAFLD (diagnosed based on imaging, blood biomarkers or liver biopsy) and incident hypertension. A meta-analytic approach might help resolve inconsistencies among previously published studies and more precisely define the nature and the magnitude of the association.
# Methods
The data of the meta-analysis are available from the corresponding author at reasonable request.
## Data sources and search strategy
We systematically searched Ovid-MEDLINE to identify articles reporting the results of longitudinal observational studies published up to March 2021 investigating the association between NAFLD and incident hypertension. The search, designed by S.C. and G.P., was performed by S.C. Articles were selected by using the terms "nonalcoholic fatty liver disease" OR "NAFLD" OR "fatty liver" OR "nonalcoholic steatohepatitis" AND "incidence" OR "new-onset" AND "hypertension" [fig_ref] Table 1: Overview of the included studies investigating the association between nonalcoholic fatty liver... [/fig_ref] , Supplemental digital content 1, http://links.lww.com/ EJGH/A725). We limited our searches to human studies without predefined language restrictions. Reference lists of included manuscripts and review articles were hand searched to identify additional studies not covered by the original database searches. The systematic review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) as outlined in [fig_ref] Table 2: Subgroup-sensitivity analyses on studies investigating the association between nonalcoholic fatty liver disease... [/fig_ref] , Supplemental digital content 1, http://links.lww.com/EJGH/A72516. Given the observational nature of the included studies, we followed the reporting items proposed by the Meta-analysis Of Observational Studies in Epidemiology for the meta-analysis of these studies [bib_ref] Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of..., Stroup [/bib_ref].
## Study selection
Only studies that met the following inclusion criteria were considered for the present systematic review and meta-analysis: (1) longitudinal design; (2) duration of follow-up ≥1 year; (3) assessment of the relationship between NAFLD and incident hypertension; (4) availability of a measure of association [hazard ratio or odds ratio (OR)] with 95% confidence intervals (CI) for the outcome of interest; (5) a diagnosis of liver steatosis based either on imaging techniques (ultrasonography, computerized tomography or transient elastography), blood/biomarkers [fatty liver index (FLI) [bib_ref] The Fatty Liver Index: a simple and accurate predictor of hepatic steatosis..., Bedogni [/bib_ref] , hepatic steatosis index [bib_ref] Hepatic steatosis index: a simple screening tool reflecting nonalcoholic fatty liver disease, Lee [/bib_ref] or other scores of liver steatosis] or liver biopsy and (6) a diagnosis of hypertension based on office blood pressure (BP) measurement by physicians or International Classification of Diseases (ICD) codes. Exclusion criteria were as follows: (1) cross-sectional studies, editorials, congress abstracts and case reports; (2) studies that did not exclude different causes of liver steatosis; (3) studies with a median follow-up <1 year; (4) studies that did not report a measure of association with 95% CI for the outcome of interest and (5) studies that were performed in the pediatric population.
## Data extraction and quality assessment
All titles and abstracts were independently examined by two investigators (S.C. and G.P.) and full-texts of potentially relevant articles were obtained and scrutinized separately by the same authors. We resolved discrepancies by consensus, referring back to the original articles. Information was extracted from all studies on study design, country, follow-up duration, the outcome of interest and covariates included in the multivariable regression models. In case of multiple publications on the same subjects, we included only the most up-to-date and comprehensive one. The risk of bias was assessed independently by two authors (S.C. and G.P.) and discrepancies were resolved by discussion. Studies were evaluated for their quality following the Newcastle-Ottawa Scale (NOS). This scale allocates a maximum of nine points for three major domains: selection of participants (maximum of four points), comparability of study groups (maximum of two points) and ascertainment of outcomes of interest (maximum of three points).
## Data synthesis and statistical analysis
Hazard ratios or ORs and corresponding 95% CI were considered as the measure of association of interest for each eligible study. We extracted the effect size from the statistical model reporting the maximum extent of adjustment for confounders. Adjusted hazard ratios and OR were pooled to calculate an overall estimate of effect size. Because we expected a relatively large heterogeneity in results, as it is a common finding when evaluating observational studies on different cohorts with varying degrees of adjustment, we used the random-effects model using the method of Der Simonian and Laird, with the estimate of heterogeneity being taken from the Mantel-Haenszel model. Statistical heterogeneity was evaluated by visual inspection of the forest plot, as well as by the Cochrane Q test and the I 2 statistics, which represents the proportion of the observed variability that cannot be explained by chance alone.
A funnel plot was constructed to evaluate the presence of publication bias by plotting the logarithm of the effect measure against the logarithm of its standard error. We also used both the Egger's test [bib_ref] Bias in meta-analysis detected by a simple, graphical test, Egger [/bib_ref] and the rank correlation Begg's test [bib_ref] Operating characteristics of a rank correlation test for publication bias, Begg [/bib_ref]. To evaluate the possible sources of heterogeneity and the robustness of our findings, we performed prespecified subgroup-sensitivity analyses by geographical location, methodology used to diagnose NAFLD and degree of covariate adjustment (with special focus on adjustment for baseline BP values and measures of adiposity). Moreover, additional sensitivity analyses were conducted to evaluate whether the pooled effect estimate was strongly influenced by a specific study. This was performed by omitting one study each time and recalculating the pooled effect estimate on the remaining studies. All statistical analyses were performed with Stata 13.0 (Stata Corp, College Station, Texas, USA). A two-tailed P value <0.05 was considered significant.
# Results
# Search results
From a total of 1108 articles identified by literature research, 1071 were excluded by title and abstract screening. We examined the full text of the remaining 37 studies. After excluding articles with a cross-sectional design or that did not report the outcome of interest (n = 18), 2 studies were not included because they reported results on the same population of two included studies, and 6 were excluded because they used different diagnostic methods to define NAFLD (mainly γ-GT levels), leading to a final number of 11 included studies that were analyzed and assessed for quality. A PRISMA flow diagram of the study selection is shown in [fig_ref] Figure 1: PRISMA flow diagram [/fig_ref].
## Features of the included articles
The main characteristics of the included studies are reported in [fig_ref] Table 1: Overview of the included studies investigating the association between nonalcoholic fatty liver... [/fig_ref]. All were observational (either prospective or retrospective) cohort studies and most of them were performed on middle-aged individuals sampled from the general population. Overall, they included 390 348 individuals (52% men) with a mean follow-up of 5.7 years (ranging from 2.6 to 9 years). Eight studies were carried out in Asia (South Korea and China), two in Europe (France and Germany) and one in the USA. Excluding one study that did not report the prevalence of NAFLD but segregated the population in FLI quartiles [bib_ref] A close relationship between non-alcoholic fatty liver disease marker and new-onset hypertension..., Roh [/bib_ref] , the mean prevalence of NAFLD was 21.5%. One study was performed only in men, while all the rest included a combined sample of men and women.
Six studies used ultrasonography to diagnose NAFLD (n = 45 924 individuals), one study used computed tomography (n = 1051 individuals) and the remaining four used the FLI (n = 343 373). Definition of hypertension was consistent in most studies and as BP ≥140/90 mmHg or the initiation of antihypertensive treatment, with one study identifying patients using ICD codes. As reported in Supplementary Table S3, Supplemental digital content 1, http://links.lww.com/EJGH/A725 4, 6 and 1, studies were considered at low (receiving at least 8 stars), medium (7 stars) and high risk of bias (<7 stars) according to NOS, respectively, thus indicating an overall low to medium risk of bias.
## Association between nonalcoholic fatty liver disease and incident hypertension
As shown in [fig_ref] Figure 2: Forest plot and pooled estimates on the effect of NAFLD on the... [/fig_ref] , the pooled hazard ratios for incident hypertension was 1.66 (95% CI, 1.38-2.01; test for overall effect z = 5.266; P < 0.001) when pooling adjusted effect estimates. The test for heterogeneity was significant (Cochran's Q = 109.85; degrees of freedom (df) = 10; P < 0.001). No study suggested a decreased risk of incident hypertension in patients with NAFLD.
When the analysis was stratified based on the methodology used to identify patients with NAFLD, the association of interest was consistent in both studies using FLI (n = 4 studies; hazard ratios 2.00; 95% CI, 1.58-2.53; test for overall effect z = 5.766; P < 0.001) and studies using imaging techniques such as ultrasonography or CT (n = 7 studies; hazard ratios 1.48; 95% CI, 1.20-1.82; test for overall effect z = 3.657; P < 0.001), with borderline heterogeneity between the two groups (Cochran's Q = 3.49; degrees of freedom (df) = 1; P = 0.062).
## Sensitivity analyses and risk of publication bias
Subgroup analyses based on follow-up duration, degree of adjustment for covariates and geographical region were performed to explore possible sources of heterogeneity and are shown in [fig_ref] Table 2: Subgroup-sensitivity analyses on studies investigating the association between nonalcoholic fatty liver disease... [/fig_ref]. Notably, an increased risk of incident hypertension in patients with NAFLD was evident in all subgroups. No significant impact was found with regards to the duration of follow-up, geographical region and adjustment for baseline BP values. On the other hand, we found that adjustment for adiposity measure at baseline (either BMI or waist circumference or both) attenuated the extent of the association. Indeed, the hazard ratio was 2.44 (95% CI, 1.84-3.22; test for overall effect z = 6.229; P < 0.001) for those not performing the adjustment (n = 4 studies) and 1.36 (95% CI, 1.20-1.54; test for overall effect z = 4.871; P < 0.001) for those performing it (n = 7 studies), with a significant between-group heterogeneity in the outcome measure (Cochrane Q = 14; df = 1; P < 0.001). No evidence of significant publication bias was found by using asymmetry analysis in the funnel plot [fig_ref] Figure 3: Funnel plot of selected studies describing the relationship between effect size and... [/fig_ref]. Furthermore, both Egger's test (P = 0.247) and rank correlation Begg's test (P = 0.312) did not show statistically significant asymmetry. Finally, sensitivity analysis (Supplementary [fig_ref] Figure 1: PRISMA flow diagram [/fig_ref] , Supplemental digital content 1, http://links.lww.com/EJGH/A725) showed that there was no significant trend suggesting that the overall result was influenced by any individual study.
# Discussion
In this large meta-analysis including 11 observational cohort studies involving 390 348 adult individuals free from hypertension at baseline from different geographical locations, we show that NAFLD is associated with a hazard ratio of 1.66 (95% CI, 1.38-2.01) for incident hypertension over a mean follow-up of 5.7 years. The extent of the association did not differ when the analysis was stratified based on diagnostic modality (blood-based panels versus imaging techniques), country of origin and adjustment for baseline BP values. On the other hand, the hazard ratio from studies that adjusted their estimates for adiposity measures (waist circumference and BMI) at baseline or at follow-up was significantly lower than that of studies that did not perform this correction, even though the association remained significant. This aspect underlies the important role of obesity as a potential confounder.
The results of the present study expand those of two previous meta-analyses focusing on the role of γ-GT as a predictor of incident hypertension [bib_ref] Gamma-glutamyltransferase and risk of hypertension: a systematic review and dose-response meta-analysis of..., Kunutsor [/bib_ref] [bib_ref] Gamma-glutamyltransferase level and risk of hypertension: a systematic review and meta-analysis, Liu [/bib_ref]. The most recent, by Kunutsor et al., [bib_ref] Gamma-glutamyltransferase and risk of hypertension: a systematic review and dose-response meta-analysis of..., Kunutsor [/bib_ref] which included 14 studies for a total of 44 582 individuals, found that in a comparison of extreme thirds of baseline γ-GT levels, the relative risk for hypertension was 1.32 (95% CI, 1.23-1.43), with the heterogeneity of estimates from different studies being explained by mean age, duration of follow-up and degree of confounder adjustment. Compared to these results, we have significantly increased the sample size (about nine times) and identified NAFLD with more accurate diagnostic methods, as γ-GT levels might be affected by a series of unrelated conditions such as alcoholic liver disease, cholestatic liver disease and induction by drugs [bib_ref] Elevated serum alanine aminotransferase and gamma-glutamyltransferase and mortality in the United States..., Ruhl [/bib_ref]. On the other hand, we cannot provide evidence on whether NAFLD severity in terms of inflammation and fibrosis impacts the magnitude of this association, as was recently suggested in a biopsy-based study involving patients with NAFLD and different degrees of histologic changes [bib_ref] Significant fibrosis predicts new-onset diabetes mellitus and arterial hypertension in patients with..., Ampuero [/bib_ref]. On this aspect, additional cohort studies of well-characterized NAFLD patients are needed.
From a pathophysiological standpoint, several mechanisms might account for the role of NAFLD as a potential driver of hypertension in the general population [bib_ref] Nonalcoholic fatty liver disease: an emerging driver of hypertension, Zhao [/bib_ref]. It is well known that liver steatosis is strongly associated with insulin resistance and hyperinsulinemia. Apart from increasing the risk of developing type 2 diabetes, insulin resistance is associated with low-grade systemic inflammation and endothelial dysfunction, which might lead to vasoconstriction. Moreover, the action of insulin on sodium handling is frequently preserved in insulin resistance and contributes to sodium retention and arterial hypertension [bib_ref] The impact of insulin resistance on the kidney and vasculature, Artunc [/bib_ref]. Other pathways linking the two conditions are represented by oxidative stress, hyperactivity of the sympathetic nervous system and the angiotensin aldosterone systems as well as increased risk of chronic kidney disease [bib_ref] Non-alcoholic fatty liver disease and risk of incident chronic kidney disease: an..., Mantovani [/bib_ref].
The current meta-analysis has several limitations that deserve to be acknowledged. First, given the observational nature of the included studies, it is not possible to definitely prove a causality link between the exposure and the outcome. Second, while most studies adjusted for several potential confounders including age, cigarette smoke and baseline BP values (as shown in [fig_ref] Table 1: Overview of the included studies investigating the association between nonalcoholic fatty liver... [/fig_ref] , the possibility of residual confounding by unmeasured factors cannot be excluded. As an example, some studies did not adjust for baseline BMI and waist circumference. It should be noted, however, that these parameters are included in the FLI equation and adjustment might therefore reduce the diagnostic ability of the score to correctly identify patients with steatosis and therefore bias results towards the null. It was therefore not possible to combine models that accounted for the same variables. While sensitivity analyses showed consistency of the association of interest independently of geographical region, most studies included Asian patients, All studies included in [fig_ref] Figure 2: Forest plot and pooled estimates on the effect of NAFLD on the... [/fig_ref] were analyzed in these subgroup analyses. a Inclusion of either BMI or waist circumference in the multivariable logistic regression model. BP, blood pressure; CI, confidence interval. who tend to develop NAFLD at lower BMI levels compared to patients of Caucasian origin and this aspect may influence the observed effect of adiposity in modulating the relationship between NAFLD and hypertension. Third, interpretation of our results demands cautiousness given the high degree of heterogeneity found between studies. While no study found a lower risk of hypertension in patients with NAFLD, variability in the magnitude of the association might result from a combination of factors including covariate adjustment, methods for NAFLD diagnosis and other potential unmeasured variables. It should also be noted that thresholds for significant alcohol consumption differed among the included studies, and not all of them systematically screened all participants for different forms of liver disease and use of steatogenic medications. More detailed analysis of heterogeneity would require pooling individual participants' data from the different studies.
Fourth, none of the included studies used a gold standard technique such as liver biopsy or magnetic resonance spectroscopy to diagnose NAFLD. In fact, while these two techniques are more reliable than both liver ultrasonography and FLI, they are expensive and time-consuming, making them unsuitable for large-scale population studies. Moreover, liver biopsy is an invasive technique with possible (although rare) life-threatening complications, raising ethical concerns related to its use in apparently healthy subjects [bib_ref] Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement, Moher [/bib_ref].
Our analysis also has some important strengths. It incorporates data from large epidemiological studies from Asia, Europe and the US including a representative pool of patients with NAFLD seen in clinical practice. Moreover, the large number of both exposed individuals and events yields high statistical power to precisely quantify the association between NAFLD and incident hypertension. Finally, there was no sign of significant publication bias affecting the analysis when evaluated by both Egger's and Begg's tests.
In conclusion, this large meta-analysis shows that NAFLD (diagnosed by either FLI or imaging techniques) is significantly associated with a ~1.7-fold increased risk of developing hypertension over a mean of 5.7 years. Moreover, obesity is an important confounder responsible for significant heterogeneity between studies and affecting the extent of this association. This underlies the need to carefully screen patients with NAFLD for the development of hypertension and the associated risk of cardiovascular events. Further studies evaluating whether NAFLD severity in terms of inflammation and fibrosis impacts on the risk of developing hypertension are needed.
[fig] Figure 1: PRISMA flow diagram. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses. [/fig]
[fig] Figure 2: Forest plot and pooled estimates on the effect of NAFLD on the risk of incident hypertension in 11 eligible studies, stratified based on the methodology used for NAFLD diagnosis. CI, confidence interval; HR, hazard ratio; HTN, hypertension; NAFLD, nonalcoholic fatty liver disease. [/fig]
[fig] Figure 3: Funnel plot of selected studies describing the relationship between effect size and standard error on the log scale. The vertical line represents the pooled effect size and the dashed lines represent the pseudo 95% confidence intervals. [/fig]
[table] Table 1: Overview of the included studies investigating the association between nonalcoholic fatty liver disease and incident hypertension [/table]
[table] Table 2: Subgroup-sensitivity analyses on studies investigating the association between nonalcoholic fatty liver disease and incident hypertension [/table]
|
Phage-inducible chromosomal islands are ubiquitous within the bacterial universe
Phage-inducible chromosomal islands (PICIs) are a recently discovered family of pathogenicity islands that contribute substantively to horizontal gene transfer, host adaptation and virulence in Gram-positive cocci. Here we report that similar elements also occur widely in Gram-negative bacteria. As with the PICIs from Gram-positive cocci, their uniqueness is defined by a constellation of features: unique and specific attachment sites, exclusive PICI genes, a phage-dependent mechanism of induction, conserved replication origin organization, convergent mechanisms of phage interference, and specific packaging of PICI DNA into phage-like infectious particles, resulting in very high transfer frequencies. We suggest that the PICIs represent two or more distinct lineages, have spread widely throughout the bacterial world, and have diverged much more slowly than their host organisms or their prophage cousins. Overall, these findings represent the discovery of a universal class of mobile genetic elements.
# Introduction
The Staphylococcus aureus pathogenicity islands (SaPIs) are a novel class of phage satellites that are intimately related to certain temperate (helper) phages, of whose life cycles they parasitize. Following infection by a helper phage or SOS induction of a helper prophage, the SaPI genome excises from the bacterial chromosome, using SaPI-encoded integrases (int) and excision functions (xis) [bib_ref] Control of Staphylococcus aureus pathogenicity island excision, Mir-Sanchis [/bib_ref] [bib_ref] Sip, an integrase protein with excision, circularization and integration activities, defines a..., Ubeda [/bib_ref] , replicates extensively using its own replicon [bib_ref] A pathogenicity island replicon in Staphylococcus aureus replicates as an unstable plasmid, Ubeda [/bib_ref] , and is efficiently packaged into infectious particles composed of phage virion proteins [bib_ref] Unravelling bacteriophage ϕ11 requirements for packaging and transfer of mobile genetic elements..., Quiles-Puchalt [/bib_ref] [bib_ref] Transducing particles of Staphylococcus aureus pathogenicity island SaPI1 are comprised of helper..., Tallent [/bib_ref] [bib_ref] Staphylococcus aureus pathogenicity island DNA is packaged in particles composed of phage..., Tormo [/bib_ref]. Some SaPIs encode capsid morphogenesis functions that remodel the phage capsid to only fit their smaller genomes [bib_ref] SaPI operon I is required for SaPI packaging and is controlled by..., Ubeda [/bib_ref] [bib_ref] Convergent evolution of pathogenicity islands in helper cos phage interference, Carpena [/bib_ref] ; others simply use fullsize phage capsids (for a recent review, see [bib_ref] The phage-inducible chromosomal islands: a family of highly evolved molecular parasites, Penadés [/bib_ref]. The hallmark of this parasitism is a key SaPI regulatory gene that diverts the phage reproduction cycle to its own end. This gene encodes a master repressor (Stl) that governs expression of the SaPI genome [bib_ref] SaPI mutations affecting replication and transfer and enabling autonomous replication in the..., Ubeda [/bib_ref]. Unlike the classical phage repressor, the SaPI Stl repressor is not cleaved following activation of the SOS response; rather the repression is lifted by the formation of a complex between the repressor and a specific helper phage protein [bib_ref] Pirating conserved phage mechanisms promotes promiscuous staphylococcal pathogenicity island transfer, Bowring [/bib_ref] [bib_ref] Phage dUTPases control transfer of virulence genes by a proto-oncogenic G protein-like..., Tormo-Más [/bib_ref] [bib_ref] Moonlighting bacteriophage proteins derepress staphylococcal pathogenicity islands, Tormo-Más [/bib_ref]. This serves to couple the SaPI life cycle with that of the helper phage, ensuring that the SaPI is not activated unless the reproductive cycle of a helper phage is in progress. Once induced, different SaPIs use different strategies to initiate specific packaging from the cognate SaPI genome [bib_ref] SaPI operon I is required for SaPI packaging and is controlled by..., Ubeda [/bib_ref] [bib_ref] Staphylococcal pathogenicity island DNA packaging system involving cos-site packaging and phage-encoded HNH..., Quiles-Puchalt [/bib_ref] , ensuring their high intra-and inter-generic transfer [bib_ref] Role of staphylococcal phage and SaPI integrase in intraand interspecies SaPI transfer, Maiques [/bib_ref] [bib_ref] Intra-and inter-generic transfer of pathogenicity island-encoded virulence genes by cos phages, Chen [/bib_ref] [bib_ref] Phage-mediated intergeneric transfer of toxin genes, Chen [/bib_ref]. Moreover, SaPIs have a huge impact on the biology of their helper phages by interfering with phage reproduction using different and complementary strategies [bib_ref] Staphylococcal pathogenicity island interference with helper phage reproduction is a paradigm of..., Ram [/bib_ref] [bib_ref] Precisely modulated pathogenicity island interference with late phage gene transcription, Ram [/bib_ref] [bib_ref] The roles of SaPI1 proteins gp7 (CpmA) and gp6 (CpmB) in capsid..., Damle [/bib_ref] ; they are also key elements driving phage evolution [bib_ref] Virus satellites drive viral evolution and ecology, Frígols [/bib_ref].
Not surprisingly, SaPI-like elements are not unique to staphylococci, and we have recently demonstrated that they are widespread in Gram-positive (GP) cocci [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref]. This new family of mobile genetic elements (MGEs), that we have called generically phage-inducible chromosomal islands (PICIs), has very well conserved features [bib_ref] The phage-inducible chromosomal islands: a family of highly evolved molecular parasites, Penadés [/bib_ref] [bib_ref] The phage-related chromosomal islands of Gram-positive bacteria, Novick [/bib_ref]. All PICIs have a conserved gene organization [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref] , and encode a pair of divergent regulatory genes, including the PICI master repressor Rpr (called Stl in the SaPIs). Left of rpr, and transcribed in the same direction, PICIs encode a small set of genes including an integrase (int) gene. Right of rpr, and transcribed in the opposite direction, the PICIs encode an excision function (xis), and a replication module consisting of a primase homolog (pri) and a replication initiator (rep), which are sometimes fused, followed by a replication origin (ori). Next to these genes, and also transcribed in the same direction, PICIs encode genes involved in phage interference, and sometimes, a terminase small subunit homolog (terS) which is responsible for the high efficiency of the SaPI packaging [bib_ref] SaPI operon I is required for SaPI packaging and is controlled by..., Ubeda [/bib_ref]. In the SaPIs, accessory genes (usually involved in virulence; [bib_ref] The phage-inducible chromosomal islands: a family of highly evolved molecular parasites, Penadés [/bib_ref] can be found either at the 3′ end of the elements or between the int and stl genes [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref].
In addition to this well-conserved gene organization, the GP PICIs have other common features that distinguish them from their cousin prophages: (i) unique attachment (att) sites that are never occupied by prophages; (ii) the absence of phage structural and lytic genes; (iii) size typically around 15 kb; (iv) sophisticated strategies to interfere with helper phage reproduction and (v) genes unique to PICIs. Orthology analyses show that PICI genes belong to large sets of orthologs, within which the first 8-10 are almost always PICI genes, most of which never appear in other genetic elements [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref] [bib_ref] The floating (Pathogenicity) Island: a genomic dessert, Novick [/bib_ref].
The preliminary analysis of the PICIs present in the GP cocci suggested that these elements could have evolved de novo in the different genera [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref] , suggesting their existence is strongly selected in nature. Therefore, and as happens with the classical MGEs, we hypothesized that these elements are likely not just confined to the GP cocci but widespread in nature. We now report the identification and characterization of a large number of PICI elements in various Gram-negative (GN) bacteria. Because they have certain key features in common and because they are so widespread, we suggest that their lifestyle may have strong selective value and that they represent a novel agency of horizontal dissemination of virulence and other important mobile genes among bacteria.
# Results
## Identification of pici candidates
Since PICIs are a successful biological strategy in GP cocci, we hypothesized that similar elements will be widespread in GN bacteria. Using the aforementioned criteria defining the PICI elements in the cocci (see material and methods for more details), we have assembled a representative collection of putative PICIs in the GN bacteria by genome searching; their genomes and characteristics are depicted in [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref] and S1, and in [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref]. Following the nomenclature proposed for the elements present in the GP bacteria, the individual PICIs are designated with reference to their speciesthus, EcCIn or PmCIn would be used for PICIs of Escherichia coli or Pasteurella multocida, respectively, where "n" would be used for the specific PICI-containing strain.
## Features of pici candidates
The putative PICIs described here have a number of features in common. (i) Occurrence: they are very common among GN bacteria, especially in members of the Enterobacteriaeae and Pastuerellaceae (Gammaproteobacteria) [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref]. (ii) Exclusivity: the KEGG orthology analyses confirmed that the GN PICI-encoded genes are PICI specific, not being present in other families of MGEs . These analyses were performed using three unrelated PICI elements, two from E. coli, EcCICFT073 and EcCIO42, and one from P. multocida, PmCIATCC43137, as representatives of the GN PICIs. (iii) Integration: in each analyzed species, the PICI att C sites are never used by temperate prophages [fig_ref] Figure 2: Characterization of GN PICI replication origins [/fig_ref] , [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref]. (iv) Transcriptional organization: contrary to the PICIs from the GP bacteria, the PICIs in the GN species are transcribed unidirectionally and rightward, unlike the vast majority of temperate phages but similar to a few (for example, phages Mu or Sp18, in E. coli O157 Sakai [bib_ref] The defective prophage pool of Escherichia coli O157: prophage-prophage interactions potentiate horizontal..., Asadulghani [/bib_ref]. This organizational differentiation suggests that the elements from GN and GP bacteria comprise at least two different lineages. (v) Replication origins: The region immediately 3′ to the replication initiation gene in the GP PICIs represents the replication origin [bib_ref] A pathogenicity island replicon in Staphylococcus aureus replicates as an unstable plasmid, Ubeda [/bib_ref]. Examination of this region for several of the PICIs from the GN organisms reveals an organization and functionality identical to that of the GP PICIs [fig_ref] Figure 2: Characterization of GN PICI replication origins [/fig_ref]. (vi) Capsid morphogenesis: some of the elements encode homologs of the PICI capsid morphogenesis genes [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref]. (vii) Interference: The two PICIs analyzed in detail show interference mechanisms similar to those previously identified in the SaPIs and in other PICI elements from Gram-positive bacteria, representing a fascinating example of convergent evolution (see below). (viii) Accessory genes: Many of the putative PICIs carry identifiable genes that do not appear to be involved in the PICI lifecycle. These accessory genes are carried exclusively by the PICIs. As noted, some are known to be involved in virulence.
## Induction and transfer
GP PICI DNA is packaged in particles composed of phage virion proteins [bib_ref] Transducing particles of Staphylococcus aureus pathogenicity island SaPI1 are comprised of helper..., Tallent [/bib_ref] [bib_ref] Staphylococcus aureus pathogenicity island DNA is packaged in particles composed of phage..., Tormo [/bib_ref]. While some GP PICIs express a small terminase subunit (TerS) homolog [bib_ref] SaPI operon I is required for SaPI packaging and is controlled by..., Ubeda [/bib_ref] , others are highly [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref] encapsidation genes are green, with the terminase small subunit gene (terS) in light green; superantigen and other virulence genes are pink; genes encoding putative phage resistance proteins are black; other accessory genes are red; genes encoding hypothetical proteins are white packaged because they carry the phage pac or cos sequence [bib_ref] Staphylococcal pathogenicity island DNA packaging system involving cos-site packaging and phage-encoded HNH..., Quiles-Puchalt [/bib_ref] [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref]. Some GP PICIs direct the formation of capsids 1/3 the size of the helper phage capsid, using different strategies [bib_ref] SaPI operon I is required for SaPI packaging and is controlled by..., Ubeda [/bib_ref] [bib_ref] Convergent evolution of pathogenicity islands in helper cos phage interference, Carpena [/bib_ref] [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref] [bib_ref] Competing scaffolding proteins determine capsid size during mobilization of Staphylococcus aureus pathogenicity..., Dearborn [/bib_ref] [bib_ref] Enterococcus faecalis prophage dynamics and contributions to pathogenic traits, Matos [/bib_ref]. Some, but not all, of the newly identified PICIs also contain a terS homolog (shown in light green); we have to date identified both the helper phages and the interfering mechanisms for two of these: EcCICFT073 and PmCI172.
## Staphylococcus aureus
## Eccicft073
Since our genomic scrutiny had revealed the presence of putative PICI-like elements in different E. coli strains, we wanted to characterize some of them in more detail. For that, we inserted a cat marker into the elements present in strains IHE3034, 18098 and O42, and a tetA marker into the element present in the CFT073 strain. The localization of the antibiotic resistant markers in the PICI elements is shown in [fig_ref] Figure 3: Phage induction of EcCICFT073 [/fig_ref]. While the two first islands encode TerS, the last two do not. The strains carrying the putative GN PICIs were treated with mitomycin C (MC) and the transfer of these elements after induction of the resident prophages was tested. As shown in [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref] , the EcCICFT073 element, but not the other three, was transferred at high frequency to the non-lysogenic DH5α (recA mutant), C600, 594 and WG5 strains. Moreover, MC induction of a CFT073 culture resulted in PICI DNA amplification, although no "PICI monomer" band appeared [fig_ref] Figure 3: Phage induction of EcCICFT073 [/fig_ref]. There was, however, a band containing covalently closed circular (CCC) DNA, indicative of PICI excision. On the basis of the following tests, we suggest that these results are indicative of helper phage induction of PICI excision, replication and packaging: we inserted a tetA marker into the chromosomal lac gene of the EcCICFT073 host strain, the uropathogenic CFT073 strain, and tested the transfer of this marker by the resident prophages. Contrary to the transfer observed for the EcCICFT073 tetA element, we did not detect transfer of the chromosomal tetA to the recipient strains, suggesting that none of the resident prophages is a generalized transducing phage. To further confirm that the transfer of EcCICFT073 tetA was PICI-specific rather than generalized transduction, we mutationally inactivated the EcCICFT073 tetA int gene and found that this abolished detectable excision and transfer [fig_ref] Figure 3: Phage induction of EcCICFT073 [/fig_ref] ; [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref]. We also mutated the pri gene and found that this greatly reduced but did not completely eliminate transfer [fig_ref] Figure 3: Phage induction of EcCICFT073 [/fig_ref] ; [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref].
We next tested for MC induction and PICI transfer with the non-lysogenic C600 derivative carrying the EcCICFT073::tetA element, and failed to detect amplification, lysis, or transfer, confirming that the EcCICFT073 transfer depends on a helper phage. Finally, we identified the EcCICFT073 helper phage. The CFT073 genome contains at least 5 prophage-like elements [bib_ref] Extensive mosaic structure revealed by the complete genome sequence of uropathogenic Escherichia..., Welch [/bib_ref] , of which #3 is actually EcCICFT073 and is listed as O-island 51. We deleted each of the other 4 and found that deletion of #4 eliminated detectable EcCICFT073::tetA induction and transfer [fig_ref] Figure 3: Phage induction of EcCICFT073 [/fig_ref] ; [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref] , whereas none of the other prophage deletions had any effect.
## Functionality of the eccicft073 genes
Having established that the EcCICFT073 island can be phage-induced, we looked in more detail at the EcCICFT073-encoded genes to confirm that these ORFs encode proteins functional in the various aspects of the PICI life cycle as previously defined for the SaPIs. There are 5 key features of the PICI life cycle: (i) excision, which may be spontaneous, (ii) replication, (iii) specific packaging, (iv) interference, and (v) induction. Of these, induction depends on the helper phage and in the GP PICIs controls all of the aforementioned features.
## Integration
To test for int (excision/circularization/integration) function, we initially performed PCR analysis with inward-and outward-directed primers as shown in [fig_ref] Figure 4: Functionality of the EcCICFT073 cos sites [/fig_ref]. Amplicons were obtained indicative of spontaneous excision and circularization, suggesting int functionality. To clearly prove this, we cloned the int-att PI segment into the pK03Blue derivative plasmids pri-ori, pri-Δori or Δpri-ori were tested for maintenance in DH5α on selective agar at the permissive (30°C) or restrictive (44°C) temperature. A One-way ANOVA with Tukey's multiple comparisons test was performed to compare mean differences within rows. Adjusted p values were as follows: pri-ori 44°C vs pri-Δori 44°C < 0.0001 **** , pri-ori 44°C vs Δpri-ori 44°C < 0.0001 **** thermosensitive pMAK700 plasmid and tested it for integration in DH5α (recA mutant) by overnight growth followed by plating on selective agar at the restrictive temperature. As controls, we generated pMAK700 derivative plasmid mutants in the int gene or in the att PI site. We obtained colonies only from plasmids carrying an intact int-att PI complex, confirming int functionality.
## Replication
The previous results suggested that the GN PICIs encode a functional replicon. We have shown previously that the Pri-Rep-ori segment of SaPIbov1, SaPI1 or EfCIV583 can drive autonomous replication of a plasmid in different GP bacteria [bib_ref] A pathogenicity island replicon in Staphylococcus aureus replicates as an unstable plasmid, Ubeda [/bib_ref] [bib_ref] SaPI mutations affecting replication and transfer and enabling autonomous replication in the..., Ubeda [/bib_ref] [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref]. Since the putative PICIs seem to have the same replicon organization [fig_ref] Figure 2: Characterization of GN PICI replication origins [/fig_ref] , we cloned the replication region of EcCICFT073, including the rep gene and the putative ori site, into the thermo-sensitive plasmid pKO3-blue [bib_ref] Genetic reductionist approach for dissecting individual roles of GGDEF proteins within the..., Solano [/bib_ref] , with rep under the control of an arabinose-inducible promoter. We also constructed plasmids carrying mutations in the rep gene or in the ori site. These plasmids were transformed into E. coli DH5α, and 5 × 10 5 cells were plated on selective agar and incubated either at the permissive (30°C) or the restrictive temperature (44°C). Strains carrying the putative PICI rep-ori replicon generated colonies in E. coli DH5α [fig_ref] Figure 2: Characterization of GN PICI replication origins [/fig_ref] , while the strains carrying the Δrep mutant plasmid did not. Note that at 44°C, the number of colonies obtained was reduced compared to 30°C, suggesting defects on plasmid segregation. Since the deletion of the ori significantly reduced the number of colonies obtained compared to the wt plasmid, these results confirm that the GN PICIs encode a functional replication module.
## Packaging
As with the phages, there are two strategies by which the SaPIs can be packaged. Some SaPIs that use the headful mechanism encode a homolog of the phage terminase small subunit (TerS S ) that directs the preferential packaging of the SaPIs [bib_ref] SaPI operon I is required for SaPI packaging and is controlled by..., Ubeda [/bib_ref]. The TerS S has the same function as the phage enzyme (TerS P ), recognizing a specific (pac) site on its cognate genome to initiate packaging. Other SaPIs carry the helper phage cos sequence in the SaPI genome, which ensures SaPI packaging depending on the helper cos phage machinery [bib_ref] Staphylococcal pathogenicity island DNA packaging system involving cos-site packaging and phage-encoded HNH..., Quiles-Puchalt [/bib_ref]. The λ cos sequence contains three regions required to interact with the packaging machinery: cosQ, cosN, and cosB. cosQ is required for the termination of chromosome packaging. cosN is the site at which terminase cuts the DNA. Initiation of DNA packaging requires cosN and the adjacent cosB site. cosB consists of three binding sites, R3, R2, and R1, for λ TerS [bib_ref] The bacteriophage DNA packaging motor, Rao [/bib_ref] [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref]. EcCICFT073 does not encode a TerS homolog, suggesting that in order to be efficiently packaged EcCICFT073 probably uses the cos packaging strategy. As shown in [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref] EcCICFT073 carries two cosN sequences that are identical to those present in its inducing phage 4 [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref]. Remarkably, these EcCICFT073 cosN sequences resemble that present in the archetypical E. coli λ and ϕ80 phages, used during decades as models in phage studies. However, we were unable to identify the phage cosB region in the EcCICFT073 element [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref] , questioning if the island carries a bona fide cos sequence. To test the EcCICFT073 and phage cos sites for function, we introduced independently the λ, ϕ80, phage 4, and each of the two EcCICFT073 cos sites (containing the putative cosQ, cosN and cosB) into plasmid pET28a, which was not transferrable by λ, and found that the cloned cos sites enabled transfer of the plasmids by phages λ and ϕ80 [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref]. Consistent with the presence of a completely different cosB sequence, transfer of the plasmids carrying the EcCICFT073 cos sites was reduced compared to that observed with the plasmids carrying the cognate phage cos sequences [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref]. The results described above raised the exciting possibility that phages λ and ϕ80 could work as helpers for EcCICFT073. To test that, and assuming EcCICFT073 is transferred using a cos mechanism, we generated one additional version of the PICI in which the antibiotic resistant marker did not change the size of the island. Note that in the EcCICFT073 tetA island the size of the element was increased by the insertion of the cassette in the middle of the island [fig_ref] Figure 3: Phage induction of EcCICFT073 [/fig_ref]. In the new variant, the cat marker replaced gene EcCICFT073 c1499 [fig_ref] Figure 3: Phage induction of EcCICFT073 [/fig_ref] , for which we have not been able to find a phenotype. The new island, generated in strain CFT073, was also introduced in the CFT073 mutant in the helper phage 4 and in the λ and ϕ80 lysogens. The different strains were SOS (MC) induced, and the transfer of the EcCICFT073 cat island analyzed. The analysis of the lysates obtained from the different CFT073 derivative strains revealed that transfer of EcCICFT073 cat by helper phage 4 was higher than that previously observed for the EcCICFT073 tetA island [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref] , supporting the idea that EcCICFT073 uses a cos mechanism for packaging. Remarkably, phages λ and 80 transferred the EcCICFT073 cat at frequencies, for phage 80, significantly higher than those observed with the helper phage 4 [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref].
We next generated a set of mutant strains in which either one of each or both cos sites present in EcCICFT073 were deleted. Deletion of the EcCICFT073 cos1 site significantly reduced transfer of the element by phages λ and ϕ80, while deletion of the EcCICFT073 cos2 site slightly increased phage ϕ80-mediated transfer of the island but significantly increased transfer mediated by phage λ; deletion of both cos sites eliminated phage-mediated transfer of the island, confirming the identity of these sequences as cos sites and that both cos sites are functional [fig_ref] Figure 4: Functionality of the EcCICFT073 cos sites [/fig_ref]. This situation mirrors SaPIbov4 [bib_ref] Adaptation of Staphylococcus aureus to ruminant and equine hosts involves SaPI-carried variants..., Viana [/bib_ref] , which also has two putative cos site in its sequence.
Finally, we generated a phage 80 terS mutant, and tested for both phage and EcCICFT073 transfer [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref]. Deletion of the phage-encoded terS gene abolished packaging and transfer of both elements, confirming that EcCICFT073 hijacks the phage proteins for transfer.
## Interference
EcCICFT073 severely interferes with phage λ reproduction [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref]. A conserved mechanism of phage interference in many PICIs is the production of PICI-sized particles that are too small to package an intact phage genome [bib_ref] Enterococcus faecalis prophage dynamics and contributions to pathogenic traits, Matos [/bib_ref] [bib_ref] Molecular genetics of SaPI1-a mobile pathogenicity island in Staphylococcus aureus, Ruzin [/bib_ref]. To test this possibility, we MC-induced the lysogenic strain carrying phage λ and EcCICFT073, the virions were pelleted and the extracted DNA was analyzed. All of the packaged DNA was phage-sized, ruling out the possibility that the observed interference was generated by capsid size redirection.
Next we searched for putative genes in the EcCICFT073 element that could be involved in this interference. EcCICFT073 encodes a putative capsid protein (ORF c1499; AAN79968) and a head decoration protein (ORF c1500; AAN79969.1). Since in SaPIbov5 the capsid protein homolog (Ccm) blocks phage reproduction by generating SaPI-sized capsids [bib_ref] Convergent evolution of pathogenicity islands in helper cos phage interference, Carpena [/bib_ref] , we analyzed whether these proteins could be involved in the observed interference using an alternative mechanism. To do that, we deleted genes c1499 or c1500 in the EcCICFT073 tetA element present in the non-lysogenic strain C600, and tested the ability of λ to generate plaques in these strains. None of the mutants allowed normal phage reproduction [fig_ref] Figure 5: EcCICFT073 interferes with λ reproduction [/fig_ref] , suggesting that other EcCICFT073 genes are responsible for the observed interference. Moreover, deletion of the c1499 or c1500 genes in the CFT073 strain did not affect the transfer of the island by the resident prophages [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref].
## Induction
The main difference between the GP and GN PICIs lies in the regulatory region [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref]. While the GP elements encode a master repressor (called Stl in the SaPIs or generically Rpr in the other GP elements) [bib_ref] SaPI mutations affecting replication and transfer and enabling autonomous replication in the..., Ubeda [/bib_ref] [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref] , the existence of a global repressor in the GN elements is unclear. One attractive candidate is AlpA, homologs of this protein being encoded by all the E. coli PICIs [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref] and S1). AlpA is predicted to act as a DNA-binding transcriptional regulator, and previous work suggested that it controls the expression of int in a P4-like phage [bib_ref] Excision of a P4-like cryptic prophage leads to Alp protease expression in..., Kirby [/bib_ref] [bib_ref] Alp suppression of Lon: dependence on the slpA gene, Trempy [/bib_ref]. We hypothesized that AlpA could act as the main repressor of the PICIs, blocking expression of the PICI genes. To test this, we initially generated different transcriptional fusions in which the lacZ reporter was fused either to EcCICFT073 ORF3 or was located after the stop codon of alpA (see scheme in . Expression of the reporter was measured in the presence or absence of alpA. Contrary to our hypothesis, AlpA is not a repressor but seems to be an activator of the EcCICFT073 genes. To confirm this, and also to know if AlpA controls int expression, we fused the promoter regions of alpA (plasmid PalpA) or int (plasmid Pint) to lacZ, using plasmid pRW224 (see scheme in . Simultaneously, the alpA gene was cloned in plasmid pBAD18 under the control of the P BAD promoter, inducible by arabinose, generating plasmid pBAD-alpA. Plasmids PalpA or Pint were introduced into the strains carrying plasmids pBAD-alpA or pBAD18, and the expression of the different promoter regions monitored in presence or absence of arabinose. As expected, AlpA activates its own expression. Contrary to previous results [bib_ref] Excision of a P4-like cryptic prophage leads to Alp protease expression in..., Kirby [/bib_ref] [bib_ref] Alp suppression of Lon: dependence on the slpA gene, Trempy [/bib_ref] , AlpA does not seem to control int expression in EcCICFT073.
Our results suggested that induction of the GN PICIs requires the expression of the AlpA activator. In support of this, inactivation of alpA significantly diminished induction and transfer of the EcCICF073 island present in the CFT073 strain [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref]. Moreover, over-expression of AlpA in the non-lysogenic E. coli C600 carrying the EcCICFT073 island resulted in induction and uncontrolled replication of the EcCICF073 element . In support of the previous results, this uncontrolled replication requires the expression of the EcCICFT073 encoded pri and int . Note that in this case, and since the strain is not lysogenic, just the closed circular (CCC) and the concatemeric (bulk) forms are observed because the element is not packaged. As occurred with the SaPIs [bib_ref] SaPI mutations affecting replication and transfer and enabling autonomous replication in the..., Ubeda [/bib_ref] , the uncontrolled replication of the EcCICFT073 affected both the growth curve and the viability of the E. coli cells .
Finally, we introduced the pJP1290 plasmid, which measures expression of alpA, in strains C600 and 594 (nonlysogenic) and the C600 and 594 derivatives lysogenic for phages λ and ϕ80, respectively. Since AlpA is absolutely required for the transfer of the island, we hypothesized that the helper phages would induce expression of alpA. . Note that alpA expression was also increased in the lysogenic strain carrying the phage ϕ80 mutant in the terS gene. Since this phage mutant can't mobilize the island, this strain was used to propose that the EcCICFT073 element hijacks the phage machinery for packaging. However, another possibility could be that terS was the EcCICFT073 inducer. The fact that this mutant induces alpA expression rule out this possibility. The identification of the phage-encoded EcCICF073 inducer is now under study.
## Pmci172
In a complementary study aimed at characterizing prophages present in the animal pathogen P. multocida, we MC-induced different lysogenic strains, the particles were pelleted and the extracted DNA was separated. As shown in [fig_ref] Figure 7: Induction of PmCI172 and PmCI86 [/fig_ref] , two bands, one of phage monomer size and the other of PICI monomer size, appeared in the analysis of strains Pm86 and Pm172. This is the classical pattern observed when a strain containing a PICI-helper phage pair is MC-induced, suggesting that P. multocida contains bona fide PICI elements, which are packaged in small capsids as are many of the analyzed PICIs [bib_ref] Convergent evolution of pathogenicity islands in helper cos phage interference, Carpena [/bib_ref] [bib_ref] Enterococcus faecalis prophage dynamics and contributions to pathogenic traits, Matos [/bib_ref] [bib_ref] Molecular genetics of SaPI1-a mobile pathogenicity island in Staphylococcus aureus, Ruzin [/bib_ref] [bib_ref] Antibiotic-induced SOS response promotes horizontal dissemination of pathogenicity island-encoded virulence factors in..., Ubeda [/bib_ref]. To confirm that this was the case, we initially sequenced both the P. multocida strains as well as the DNA obtained from the induced lysates. As expected, we identified a PICI element in both strains, which incidentally were identical [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref]. While two phages were present in strain Pm86 [fig_ref] Figure 7: Induction of PmCI172 and PmCI86 [/fig_ref] , strain Pm172 only contained one [fig_ref] Figure 7: Induction of PmCI172 and PmCI86 [/fig_ref]. The latter phage was also present in strain Pm86, suggesting this is the helper phage for the PmCI172 and PmCI86 PICIs. This phage belongs to the Mu family of phages, demonstrating for the first time a Mu-like phage, belonging to the Myoviridae family, as helper phage for a PICI element. Next, we confirmed that the PmCI172-and Pm86-Mu phage are a bona fide PICI-helper phage interaction by performing a Southern blot analysis of the packaged DNA observed in [fig_ref] Figure 7: Induction of PmCI172 and PmCI86 [/fig_ref] , using PmCI172-and phage Mu-specific probes. Southern blot analyses revealed that the Mu phage probe hybridized both with the PICI-sized (small) and the phage-sized (large) DNA bands [fig_ref] Figure 7: Induction of PmCI172 and PmCI86 [/fig_ref] , suggesting that the phage DNA can be packaged in the PmCI-sized particles. Packaging of a significant proportion of phage DNA in the small particles generates defective phages, confirming that PmCI172 and PmCI86 PICIs interfere with phage reproduction using a very well conserved strategy, production of PICI-sized capsids, which is a key feature of the previously analyzed GP PICI elements [bib_ref] Convergent evolution of pathogenicity islands in helper cos phage interference, Carpena [/bib_ref] [bib_ref] Enterococcus faecalis prophage dynamics and contributions to pathogenic traits, Matos [/bib_ref] [bib_ref] Molecular genetics of SaPI1-a mobile pathogenicity island in Staphylococcus aureus, Ruzin [/bib_ref] [bib_ref] Antibiotic-induced SOS response promotes horizontal dissemination of pathogenicity island-encoded virulence factors in..., Ubeda [/bib_ref].
Finally, to confirm the existence of the PICI-sized capsids, we subjected the virions to electron microscopy (EM). The phage-sized particles had the characteristic size and shape of this class of bacteriophages: hexagonal capsids of approximately 59 to 63 nm length and 51 to 56 nm width and long contractile tails, which are approximately 150 nm long and 15 to 16 nm wide. [fig_ref] Figure 7: Induction of PmCI172 and PmCI86 [/fig_ref]. As expected, the majority of the particles had small, hexagonal heads, about 38-39 nm length and 37 nm wide, attached to approximately 150 nm long and 15 nm wide tail [fig_ref] Figure 7: Induction of PmCI172 and PmCI86 [/fig_ref]. This result confirms that PmCI172 causes the formation of small capsids, consistent with its smaller genome size. As occurs with SaPIbov5, which encodes a capsid protein responsible of the production of the SaPI-sized small capsids [bib_ref] Convergent evolution of pathogenicity islands in helper cos phage interference, Carpena [/bib_ref] , PmCI172 also encodes a major capsid protein homolog [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref] , which could be involved in this process. Unfortunately, the absence of tools to manipulate P. multocida has impaired us deciphering in detail the biology of the PmCI172 element.
# Discussion
In this report we demonstrate that PICIs, of which the staphylococcal SaPIs are a subset, are not just confined to GP Characterization of the EcCICFT073 alpA gene. a-c Schematic representation of the different blaZ transcriptional fusions generated. The relevant genes are shown. b-d Miller assay performed using the plasmids represented in (a-c). Results (average ± s.d.) of three independent assays are shown. A 2-way ANOVA with Tukey's multiple comparisons test was performed to compare mean differences within rows. Adjusted p values were as follows: pJP1290 vs pJP1291 = 0. 0011 ** , pJP1288 vs pJP1289 = 0.0001 *** , PalpA < 0.0001 **** , Pint > 0.9999 ns . ns, not significant. e EcCICFT073 excision and replication after expression of the cloned alpA gene. A non-lysogenic derivative of strain C600 carrying EcCICFT073 was complemented with plasmid pBAD18 (empty plasmid) or plasmid pJP2037, which carries alpA under the control of the P BAD promoter. As controls, non-lysogenic derivatives of strain C600 carrying EcCICFT073 Δint or EcCICFT073 Δpri were complemented with plasmid pJP2037. One milliliter of each culture (optical density (OD) 600nm = 0.3) was collected 2 h after treatment with 0.02% arabinose and used to prepare standard minilysates, which were resolved on a 0.7% agarose gel, Southern blotted and probed for EcCICFT073 DNA. In these experiments, because no helper phage is present, the excised EcCICFT073 DNA appears as covalently closed circular molecules (CCC). f Helper phages activate alpA transcription. Different lysogenic and non-lysogenic strains, containing plasmid pJP1290, were MC-induced and assayed for β-galactosidase activity. Results (average ± s.d.) of three independent assays are shown. Strains: C600 (non-lysogenic), JP10400 (C600 lysogenic for phage λ), 594 (non-lysogenic), JP12507 (594 lysogenic for phage 80) and JP15151 (594 lysogenic for phage 80 ΔterS). A 2-way ANOVA with Tukey's multiple comparisons test was performed to compare mean differences within rows. Adjusted p values were as follows; C600 > 0.9999 ns , Phage Lambda < 0.0001 **** , 594 > 0.9999 ns , Phage 80 < 0.0001 **** , Phage 80 ΔterS < 0.0001 **** . ns, not significant bacteria but are widespread among other bacteria. Although all the GN PICIs identified in this study are in Gammaproteobacteria, and more specifically in members of the Enterobacteriaeae and Pastuerellaceae, the fact that similar elements can be found both in GP and GN bacteria confirms the universality of this novel family of MGEs. PICIs are phage satellites with a unique genomic organization which is easy to see in KEGG genome maps and sets them apart from other mobile genetic elements. They interact with certain "helper" bacteriophages in a unique and highly characteristic manner: in the GP cocci, the PICI encode an SOS-insensitive master repressor, Stl, that maintains it in a b Southern blot of the DNA shown in panel (a), using a phage-or PmCI-specific probe. c Electron microscopy analysis of the MC-induced P. multocida 172 lysate. Several different fields are shown, containing normal Mu phage particles and PmCI172 particles (arrows), which have smaller heads. Scale bars are 50 nm quiescent state and is counteracted by a specific phage protein, initiating the PICI life cycle [bib_ref] Moonlighting bacteriophage proteins derepress staphylococcal pathogenicity islands, Tormo-Más [/bib_ref] [bib_ref] Phage-inducible islands in the Gram-positive cocci, Martínez-Rubio [/bib_ref]. In the GN bacteria, it seems the PICIs use a completely different strategy coupling the phage and the PICI cycles. In this case, the GN PICIs encode AlpA, an activator of the PICI cycle whose expression depends on the helper phage. Why GP and GN bacteria use different strategies to be induced is a mystery. In both cases, however, PICI induction is followed by PICI replication, encapsidation of PICI DNA in infectious particles composed of phage proteins and finally high transfer of these elements.
Like their GP counterparts, GN PICIs also severely interfere with helper phage reproduction, this probably being one of the raisons d'être of these elements. The PICI genomes are usually about 1/3 the size of their helper phage genomes and among the GP PICIs, many encode proteins that redirect the helper phage to form capsids 1/3 the size of the phage capsid [bib_ref] SaPI operon I is required for SaPI packaging and is controlled by..., Ubeda [/bib_ref] [bib_ref] Convergent evolution of pathogenicity islands in helper cos phage interference, Carpena [/bib_ref] [bib_ref] Enterococcus faecalis prophage dynamics and contributions to pathogenic traits, Matos [/bib_ref]. This strategy is also used by the GN PICIs. Although we have not identified the genes responsible for this process, none of the proteins encoded by the PmCI172 element resemble previously characterized proteins, suggesting that capsid size redirection is a convergent evolutionary process characteristic of all the PICI elements, which in turn defines the size of these elements.
As occurs with GP PICIs, GN PICIs have an important role in virulence and several of the newly identified PICIs present in GN bacteria contain recognizable accessory genes. These are listed in [fig_ref] Table 1: Transductants/ml of lysate, using E [/fig_ref]. The identities of many of these genes are based on annotations; just a few of them have been tested experimentally. Several elements, including islands harbored by variants of the notorious E. coli O157:H7 encode RgdR, which regulates expression of the type III secretion (T3S) system, encoded by the locus of enterocyte effacement (LEE) of the E. coli pathogenicity island, PAI-I. Deletion of this island in E. coli O157 strain TUV93-0 resulted in a reduction in LEE expression and T3S [bib_ref] Identification of a novel prophage regulator in Escherichia coli controlling the expression..., Flockhart [/bib_ref]. This deletion also reduced the capacity of the bacteria to attach to epithelial cells and reduced the shedding of E. coli O157 by sheep. RgdR also controls the expression of other genes, including those involved in motility, suggesting a global role in E. coli gene expression [bib_ref] Identification of a novel prophage regulator in Escherichia coli controlling the expression..., Flockhart [/bib_ref]. It is especially relevant for this report that EcCICFT073, and the related EcCI042, two of the rgdRcontaining elements in E. coli, correspond to O-island 51, a 15 kb O157:H7 island annotated, as usual, as a defective prophage [bib_ref] Identification of a novel prophage regulator in Escherichia coli controlling the expression..., Flockhart [/bib_ref]. In addition to RgdR, which is encoded by all the analyzed E. coli PICIs, some E. coli PICIs encode allelic variants of PerC. Remarkably, some but not all the PICIcoded PerC homologs also activate expression of the LEE pathogenicity island [bib_ref] The LEE1 promoters from both enteropathogenic and enterohemorrhagic Escherichia coli can be..., Porter [/bib_ref].
We suggest that all of the PICIs in the GN species are coancestral and represent a separate but very similar lineage, since their transcriptional organization is mirrored in rare phages from GN organisms. At the same time, the unique organization of the replication origin is the same for both the GP and GN lineages, which is consistent with coancestry. Although this organization was not a criterion for the identification of the PICIs, it is shared by all those identified in this study and by no other known bacterial replicon. Coancestry could mean that the divergent evolution of PICIs and host chromosomes has simply been parallel, or that the PICIs have been horizontally transferred and have retained greater similarities than their host bacteria. Current studies in our labs are focused on deciphering these interesting evolutionary questions. Our overall view, therefore, is that the PICIs represent two or more families of unique mobile elements whose biological features have been strongly selected during long-term evolution, so that they represent coherent clades that have diverged from and evolved in parallel with temperate phages (or proto-phages) and have spread widely among diverse bacterial genera. If this spread has been via helper phage induction and packaging, then it is likely that spread would have occurred between species that share phage adsorption receptors, leading to a chain of island-hopping (or rather host-hopping) events that could, in principle, be tracked by comparing PICI gene similarities as a function of the relatedness of host bacterial phage receptors.
The P2-P4 system in E. coli represents a theoretical precedent for phage satellites, in which P4 utilizes P2 virion proteins to form small capsids into which its small genome is packaged [bib_ref] Mechanisms of genome propagation and helper exploitation by satellite phage P4, Lindqvist [/bib_ref] [bib_ref] Pirates of the Caudovirales, Christie [/bib_ref]. PICIs and P4 share some features in common: (i) phage-mediated activation of their life cycles, which in the case of P4 depends on the P2 Cox protein [bib_ref] Activation of prophage P4 by the P2 Cox protein and the sites..., Saha [/bib_ref] , while for the GN PICIs the activator remains to be identified; (ii) autonomous replication, which in the case of P4 depends on the α protein [bib_ref] Bacteriophage P4 DNA replication, Ziegelin [/bib_ref] and its cognate ori site (named ccr [bib_ref] DNA replication in phage P4: characterization of replicon II, Magnoni [/bib_ref] [bib_ref] Identification of two replicons in phage-plasmid P4, Tocchetti [/bib_ref] ; (iii) carriage of the helper phage cos sequence in their genomes, which allows these elements to parasite the packaging machinery of their cognate helper phages; (iv) interference with phage reproduction, a process that in P4 depends on Sid, a protein that redirects capsid formation generating P4-sized particles [bib_ref] Regulation of icosahedral virion capsid size by the in vivo activity of..., Agarwal [/bib_ref].
However, and in spite of these similarities, P4 and the PICIs present in E. coli show important differences supporting the idea that they have evolved independently: (i) P4 and PICIs are fundamentally different both in genomic structure (see and in gene content. As occurred with the PICIs, the KEGG analysis revealed that the P4 genes are exclusive of this family of elements . Importantly, only three PICI-coded proteins (AlpA, Pri-Rep and Int) showed distant homology with the P4-coded proteins, corresponding to proteins that have similar roles in both elements. We have identified 4 allelic variants of AlpA in the E. coli PICIs , all of them being remotely related to the AlpA-homolog (named Vis) encoded by P4 . In fact, these proteins seem to have a different role in each of the elements: while Vis is required for P4 excision [bib_ref] Bacteriophage P4 Vis protein is needed for prophage excision, Calì [/bib_ref] , AlpA controls induction of the EcCICFT073 cycle. In the case of the Pri-Rep proteins, a similar scenario is observed: 4 allelic variants are found in the PICI elements from E. coli [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref] , which are different in sequence from the α protein of P4 [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref]. The situation with the Int proteins is even more interesting. We have identified in E. coli 36 att C sites for phages, 8 for PICIs and 4 for P4 elements [fig_ref] Figure 2: Characterization of GN PICI replication origins [/fig_ref]. All the att C sites are exclusive to one type of element except one, which is shared by PICIs and P4. Importantly, and with the exception of the int gene, the P4 and PICI elements that integrate in the same att C conserved their own features and are completely different both in sequence and structure [fig_ref] Figure 1: Genome maps for PICIs [/fig_ref]. (ii) PICIs do not have a plasmid state, nor encode genes to control plasmid copy number or plasmid segregation. (iii) PICIs play an important role in virulence, with many E. coli elements encoding two activators of the LEE pathogenicity island. (iv) P2 and P4 mutually cross induce one another through back-andforth transcriptional activation. Indeed, P4 forms plaques on a P2 lysogen, but these are P2 plaques resulting from P4 induction of P2, whereas we have not been able to see phage induction mediated by an incoming GN PICI. (v) In E. coli, P4 and PICI elements seem to parasitize two completely different families of helper phages: Myoviridae and Siphoviridae, respectively. Taken together, all these similarities and divergences highlight the idea that although these elements have evolved independently, their parallel modus vivendi has been strongly selected in nature.
A similar scenario occurs with a family of pathogenicity islands recently identified in Vibrio cholerae. These elements, denominated PLEs for phage-inducible chromosomal island-like elements, are widespread genomic islands that share some features with the classical PICIs, including mobilization by helper phages and severe interference with phage reproduction [bib_ref] A highly specific phage defense system is a conserved feature of the..., O'hara [/bib_ref] [bib_ref] A bacteriophage encodes its own CRISPR/Cas adaptive response to evade host innate..., Seed [/bib_ref]. In contrast to the GN PICIs, PLEs are larger in size and have both different genetic organization and gene content. Thus, although they encode an integrase, they do not encode identifiable regulatory, replication or packaging modules, even though infection by a helper phage leads to PLE excision, replication and packaging [bib_ref] A highly specific phage defense system is a conserved feature of the..., O'hara [/bib_ref]. Whether these functions depend on the helper phage machinery, or by contrast are provided by still-unidentified PLE genes remains to be determined. Further studies are required to understand the origins of, and relationships among, these subcellular elements and their cognate phages.
Although cos sites were long ago cloned into E. coli plasmids, generating cosmids, which are widely used as cloning vectors owing to their efficient transfer by phage λ, our results represent the first demonstration of naturallyoccurring phage λ-mediated transfer. Since SaPIbov5 is also transferred by helper cos phages, our results with the PICIs represent a paradigm shift involving cos phages in gene transfer. The PICIs are a special case because they undergo phage-like replication, generating concatemeric postreplicative DNA, which would have cos sites spaced one genome apart. However, EcCICFT073 lacks the capsid morphogenesis genes carried by most SaPIs and is therefore packaged in full-sized phage capsids, which are designed to accommodate the 45 kb phage genomes. This would mean that the phage terminase skips two successive cos sites during packaging; the mechanism by which this occurs remains to be determined, although it is likely that the expansion generated once the capsid is fulfilled controls this phenomenon.
Bacteria are successful as commensal organisms or pathogens in part because they adapt rapidly to selective pressures. MGEs play a central role in this adaptation process and are a means to transfer genetic information among and within bacterial species. Here, and as is the case with other classical MGEs like plasmids, transposon or phages, we demonstrate that the PICIs are widespread in nature. These findings represent the discovery of a universal agency of horizontal dissemination of important accessory genes in the bacterial universe.
# Materials and methods
## Identification of pici candidates
The following criteria were initially used to identify putative PICIs in the GN genomes: (i) well-conserved gene organization, including an integrase, a replicating module and a master transcriptional regulator; (ii) unique attachment (att) sites that are never occupied by prophages; (iii) the absence of phage structural and lytic genes; and (iv) size typically around 12-15 kb. After this initial screening, the analysis of orthologies points to elements that might correspond to PICIs. Examination of the corresponding KEGG genome maps (http://www.genome.jp/kegg; release May 1 2016) was used to confirm the identifications. In these analyses we examined the E. coli and the P. multocida genomes that have been coded for KEGG because the KEGG genome maps enable PICI-like elements to be readily identified.
## Bacterial strains and growth conditions
Bacterial strains used in these studies are listed in Supplementary . E. coli strains were grown at 37°C or 30°C overnight on Luria-Bertani agar and on Luria-Bertani broth with shaking (180 r.p.m.). P. multocida strains were grown at 37°C overnight on Brain-Heart-Infufion agar and on Brain-Heart-Infufion broth with shaking (250 r.p.m.). Ampicillin (100 µg ml −1 ), Kanamycin (30 µg ml −1 ), Chloramphenicol (20 µg ml −1 ) or Tetracycline (20 µg ml −1 ), from Sigma, were added when appropriate. Phage and PICI analyses were performed, as described previously [bib_ref] Pirating conserved phage mechanisms promotes promiscuous staphylococcal pathogenicity island transfer, Bowring [/bib_ref] [bib_ref] Phage dUTPases control transfer of virulence genes by a proto-oncogenic G protein-like..., Tormo-Más [/bib_ref] [bib_ref] Moonlighting bacteriophage proteins derepress staphylococcal pathogenicity islands, Tormo-Más [/bib_ref].
# Dna methods
Gene insertion was performed as described [bib_ref] One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products, Datsenko [/bib_ref]. The tetracycline resistance marker (tetA) or chloramphenicol resistance marker (cat) was amplified by PCR with primers listed in . The resulting PCR products were used to transform the recipient strain harboring plasmid pKD46 [bib_ref] One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products, Datsenko [/bib_ref] , which expresses Lambda Red recombinase. PCR was performed to verify the insertion of markers.
## Plasmid construction
Plasmids used in this study were constructed by cloning PCR products amplified with oligonucleotide primers purchased from Invitrogen. Clones were sequenced by Eurofins. Plasmids and primers are listed in .
## Southern blot
Following plasmid (arabinose) or phage (MC) induction, samples were taken and pelleted at the indicated times. Samples were re-suspended in 50 μl lysis buffer (47.5 μl TES-Sucrose and 2.5 μl lysozyme [10 μg/ml]) and incubated at 37°C for 1 h. Subsequently, 55 μl of SDS 2% proteinase K buffer (47.25 μl H2O, 5.25 μl SDS 20%, 2.5 μl proteinase K [20 mg/ml]) was added and incubated at 55°C for 30 min. Samples were vortexed with 10 μl of 10 × loading dye. Cycles of incubation in dry ice with ethanol and at 65°C were performed. Chromosomal DNA was separated by agarose gel electrophoresis. Samples were run on 0.7% agarose gel at 30 V overnight. Nylon membranes (Hybond-N 0.45 mm pore size filters; Amersham Life Science) were used for the transfer of DNA. DNA was detected using a DIG-labeled probe and anti-DIG antibody.
## Β-galactosidase assays
Strains were grown at 37°C in LB medium containing the appropriate antibiotics. Following plasmid (arabinose) or phage induction (MC), samples were taken and pelleted at the indicated times. The Miller methodwas used to measure β-galactosidase levels. The average of at least three independent experiments is shown in Miller units.
Electron microscopy (P. multocida) Ten milliliters of filtered P. multocida Pm172 phage lysate were centrifuged at 40,000 × g for 90 min at 4°C. The supernatant was carefully removed and the pellet was resuspended gently with 0.5 ml of 0.1 M ammonium acetate (pH 7.3). The suspension was stored overnight at 4°C for negative staining and visualization by TEM. Three hundredmesh carbon-coated nickel grids were dropped onto 50-100 µL of phage suspension. The grids were allowed to adsorb phage suspension for 2 min. The grids were washed 3 times with dH 2 O for 10″, excess fluid was removed using Whatman filter paper and the grids were negatively stained with 2% ammonium molybdate for 30″. Excess staining solution was removed with Whatman filter paper. The grids were allowed to dry at room temperature for~15-20 min. The samples were examined by TEM (FEI Tecnai TF20) at 200 kV using Gatan Microscopy Suite Software.
[fig] Figure 1: Genome maps for PICIs. Genomes are aligned according to the prophage convention, with the integrase gene (int) at the left end. Genes are colored according to their sequence and function: int is yellow; transcription regulator (stl or rpr (GP), and alpA or merR (GN)) is dark blue; replication genes are purple; [/fig]
[fig] Figure 2: Characterization of GN PICI replication origins. a Comparative map of the replication origins of several E. coli PICIs. The iterons are represented by arrows, and their sequences are shown at left. Note that there are always two sets of iterons flanking an AT rich region, which could be the melting site. b Testing EcCICFT073 pri-rep-ori function. [/fig]
[fig] Figure 3: Phage induction of EcCICFT073. CFT073 strains were SOS induced with MC (2 μg/ml). Samples were removed at the indicated time points and used to prepare minilysates, which were resolved on an 0.7% agarose gel (upper panel), and Southern blotted (lower panel) with an EcCICFT073 probe. M: Southern blot molecular marker (DNA molecular weight marker VII; Roche) [/fig]
[fig] Figure 4: Functionality of the EcCICFT073 cos sites. EcCICFT073 cat wt, or its derivatives carrying mutations in the cos1 or/and cos2 sites, were introduced in the lysogenic strains for phages λ or ϕ80. The different strains were MC induced (2 μg/ml) and the transfer of the island quantified. A t-test was performed to compare the wt against the different cos mutants. Adjusted p values were as follows: Phage Lambda EcCICFT073 Δcos1 = 0.002 ** , Phage Lambda EcCICFT073Δcos2 = 0.0266 * , Phage 80 EcCICFT073 Δcos1 = 0.003 *** , Phage ϕ80 EcCICFT073 Δcos2 = 0.0002 *** [/fig]
[fig] Figure 5: EcCICFT073 interferes with λ reproduction. E. coli strains C600, JP12677 (C600 EcCICFT073 tetA-positive), JP15181 (C600 EcCICFT073 tetA-positive Δc1499) or JP15182 (C600 EcCICFT073 tetA-positive Δc1500) were infected with phage λ (∼500 pfu per plate), plated on phage bottom agar, and incubated for 24 h at 37°C. Plates were stained with 0.1% TTC in LB and photographed genes [/fig]
[fig] Figure 7: Induction of PmCI172 and PmCI86. a DNA extracted from a phage lysates of MC-treated cultures of P. multocida strains 172 or 86. [/fig]
[table] Table 1: Transductants/ml of lysate, using E. coli C600, WG5, 594 or DH5α (recA mutant) as recipient strains. The means of results from three independent experiments are presented. Variation was within ± 5% in all cases Phages deleted in strain E. coli CFT073 [/table]
|
Clinical outcome of breast cancer in carriers of BRCA1 and BRCA2 mutations according to molecular subtypes
olivier trédan 1 & S.intidhar Labidi-Galy 5,11 ✉ BRCA1/BRCA2 genes play a central role in DnA repair and their mutations increase sensitivity to DNA-damaging agents. There are conflicting data regarding the prognostic value of BRCA germline mutations in breast cancer (Bc) patients. We collected clinical, pathological and genetic data of a cohort 925 BC patients preselected for genetic screening and treated with neoadjuvant or adjuvant chemotherapy, of whom 266 were BRCA carriers. Overall, 171 women carried a BRCA1 mutation, 95 carried a BRCA2 mutation, and 659 were non-carriers. In the entire cohort, there was a prolonged disease-free survival (DfS) for BRCA carriers (hazard ratio (HR) = 0.63; 95% confidence interval (CI), 0.44-0.90 for BRCA1; HR = 0.72; 95%CI, 0.47-1.1 for BRCA2; p = 0.020) and a trend toward prolonged disease-specific survival (DSS; HR = 0.65; 95%CI, 0.40-1.1 for BRCA1; HR = 0.78; 95%CI, 0.44-1.38 for BRCA2; p = 0.19) though not statistically significant. In the TNBC group, BRCA carriers had prolonged DFS (adjusted HR = 0.50; 95%CI, 0.28-0.89 for BRCA1; adjusted HR = 0.37; 95%CI, 0.11-1.25, for BRCA2; p = 0.034) and DSS (adjusted HR = 0.42; 95%CI, 0.21-0.82 for BRCA1; adjusted HR = 0.45; 95%CI, 0.11-1.9 for BRCA2; p = 0.023). In the non-TNBC group, the BRCA1 or BRCA2 mutations did not have any impact on survival. these results suggest that BRCA1/BRCA2 germline mutations are associated with prolonged survival only if women were diagnosed with TNBC.BRCA1/BRCA2 germline mutations account for approximately 5% of all breast cancers 1 . These tumor suppressor genes encode large, ubiquitous and multifunctional proteins that play a central role in DNA repair, cell-cycle control and chromosomal stability 2 . Cells with non-functional BRCA1/BRCA2 proteins are severely impaired in their ability to repair DNA double strand breaks (DSBs) through homologous recombination 2 . As a consequence, tumors harboring deleterious mutations of BRCA1/BRCA2 genes are highly sensitive to DNA-damaging agents, such as interstrand crosslinking agents (platinum or alkylating agents), topo-isomerase II inhibitors (anthracyclines) or PARP inhibitors 2-4 .In breast cancer patients, the tumor phenotype differs according to the BRCA1 or BRCA2 germline mutation status. BRCA1 mutation carriers mainly develop triple negative breast cancers (TNBC), whereas BRCA2 carriers are more likely to develop estrogen receptor (ER) and/or progesterone receptor (PR) positive tumors 5 . Not all BRCA carriers who develop breast cancer receive adjuvant chemotherapy, depending on several factors, including
tumor stage, grade and molecular subtype. Currently, there are conflicting data regarding the predictive and prognostic values of BRCA mutations on the survival of non-metastatic breast cancer patients [bib_ref] Clinical outcomes of breast cancer in carriers of BRCA1 and BRCA2 mutations, Rennert [/bib_ref] [bib_ref] Breast cancer prognosis in BRCA1 and BRCA2 mutation carriers: an International Prospective..., Goodwin [/bib_ref] [bib_ref] Effects of BRCA1-and BRCA2-related mutations on ovarian and breast cancer survival: a..., Zhong [/bib_ref]. BRCA carriers with TNBC have been shown to be more sensitive to DNA-damaging agents 9-15 but this did not translate into a survival benefit [bib_ref] Clinical outcomes of breast cancer in carriers of BRCA1 and BRCA2 mutations, Rennert [/bib_ref] [bib_ref] Outcome of triple negative breast cancer: comparison of sporadic and BRCA1-associated cancers, Tung [/bib_ref] [bib_ref] Prevalence of BRCA1 mutations and responses to neoadjuvant chemotherapy among BRCA1 carriers..., Wang [/bib_ref] [bib_ref] Outcome of triple-negative breast cancer in patients with or without deleterious BRCA..., Bayraktar [/bib_ref] [bib_ref] Neo-adjuvant doxorubicin and cyclophosphamide followed by paclitaxel in triple-negative breast cancer among..., Paluch-Shimon [/bib_ref].
BRCA germline mutations account for approximately 10-15% of ovarian cancers [bib_ref] BRCA mutation frequency and patterns of treatment response in BRCA mutation-positive women..., Alsop [/bib_ref]. The majority of ovarian cancers that develop in BRCA carriers (either BRCA1 or BRCA2) are high-grade serous ovarian carcinomas (HGSOC). Ovarian cancers are frequently diagnosed at advanced stages and receive platinum-based chemotherapy. Several studies have shown that among ovarian cancer patients, BRCA1 and especially BRCA2 carriers respond better than non-carriers to platinum-based chemotherapy and have prolonged survival [bib_ref] Association of BRCA1 and BRCA2 mutations with survival, chemotherapy sensitivity, and gene..., Yang [/bib_ref] [bib_ref] Association between BRCA1 and BRCA2 mutations and survival in women with invasive..., Bolton [/bib_ref] [bib_ref] Inherited Mutations in Women With Ovarian Carcinoma, Norquist [/bib_ref]. We hypothesized that BRCA germline mutations would lead to prolonged survival in breast cancer patients treated by DNA-damage agents such as alkylating agents and/or anthracylines [bib_ref] TBCRC 031: Randomized Phase II Study of Neoadjuvant Cisplatin Versus Doxorubicin-Cyclophosphamide in..., Tung [/bib_ref]. We conducted a multicentric retrospective study with the primary objective of assessing the prognostic value of BRCA germline mutation on survival among stage I-III breast cancer patients treated with chemotherapy. Patients were included if they have been selected for genetic testing of BRCA germline mutation.
Results patient demographics and clinical characteristics. From the entire cohort, a total of 925 patients were identified (677 from the French cohort and 248 from the Swiss cohort)(supplementary [fig_ref] Figure 1: DFS and DSS according to BRCA1/BRCA2 status and molecular phenotype [/fig_ref] , of whom 659 were non-carriers, 171 were BRCA1 carriers, and 95 were BRCA2 carriers (supplementary . Patient demographics, tumor characteristics, and type of administered chemotherapy are summarized in . The median age at diagnosis was similar between carriers and non-carriers. Most BRCA1 carriers developed TNBC (68%) compared to 19% among BRCA2 carriers and 24% among the non-carriers (p < 0.0001). BRCA1 carriers were more likely to develop high grade (p < 0.0001) and high mitotic index tumors (p < 0.0001). Axillary node involvement was more frequent in BRCA2 carriers (p = 0.016).
ER, PR, and HER-2 status were available for 858 patients. Among the 270 who developed TNBC, 106 were BRCA1 carriers, 16 were BRCA2 and 148 were non-carriers. Patients and tumor characteristics were comparable between BRCA carriers and non-carriers (supplementary. Among the 588 women who developed non-TNBC, BRCA1 carriers were older than BRCA2 and non-carriers (p = 0.014; supplementary [fig_ref] Table 3: Multivariate analysis of DFS and DSS in TNBC [/fig_ref]. BRCA1 carriers developed tumors displaying higher grade (p = 0.056), and a higher mitotic index (p = 0.047) and were less frequently expressing ER (p = 0.0053) than BRCA2 carriers or non-carriers. HER-2 was less frequently overexpressed/amplified in tumors from BRCA carriers compared to non-carriers (p = 0.004). chemotherapy. The majority of patients received adjuvant chemotherapy (72% for the entire cohort, 66% for TNBC, and 73% for non-TNBC). Most of the patients received two DNA damaging-agents: an alkylating agent (95%) and an anthracycline (82%; . Non-carriers were more likely to receive taxanes (p = 0.046; , in particular among those who developed TNBC (p = 0.0088; supplementary. Non-carriers more frequently received trastuzumab (p < 0.0001; . Very few patients received platinum derivatives (3%; .
Survival estimates. The median follow-up for the entire cohort was 7.3 years Subgroup analysis by molecular subtype revealed that BRCA carriers had significantly prolonged DFS and DSS in the TNBC subgroup only (n = 270; [fig_ref] Table 3: Multivariate analysis of DFS and DSS in TNBC [/fig_ref]. After adjustment for nodal status, BRCA1 (5-year rate 91%; HR = 0.50; 95%CI, 0.28-0.89) and BRCA2 carriers (5-year rate 93%; HR = 0.37; 95%CI, 0.11-1.25) had prolonged DFS compared to non-carriers (5-year rate 77%; p = 0.034; [fig_ref] Table 3: Multivariate analysis of DFS and DSS in TNBC [/fig_ref] and [fig_ref] Figure 1: DFS and DSS according to BRCA1/BRCA2 status and molecular phenotype [/fig_ref]. BRCA1 (5-year rate 91%; HR = 0.42; 95%CI, 0.21-0.82) and BRCA2 carriers (5-year rate 93%; HR = 0.45; 95%CI, 0.11-1.9) consistently had prolonged DSS compared to non-carriers (5-year rate 79%; p = 0.023; [fig_ref] Table 3: Multivariate analysis of DFS and DSS in TNBC [/fig_ref] and [fig_ref] Figure 1: DFS and DSS according to BRCA1/BRCA2 status and molecular phenotype [/fig_ref]. The landmark analysis at one year performed as a sensitivity analysis was consistent with this estimated impact of BRCA status on DFS and DSS in the TNBC" (supplementary [fig_ref] Table 4: Pathologic complete response according to BRCA status and molecular subtype [/fig_ref]. For women with non-TNBC, the BRCA1 (5-year rate 91%; HR = 0.91; 95%CI, 0.50-1.7) or BRCA2 (5-year rate 87%; HR = 1.1; 95%CI, 0.70-1.9) status did not have any impact on DFS (p = 0.88; supplementary and [fig_ref] Figure 1: DFS and DSS according to BRCA1/BRCA2 status and molecular phenotype [/fig_ref]. Similarly, the BRCA1/BRCA2 status did not have any impact on the 5-year DSS (p = 0.93; supplementary and [fig_ref] Figure 1: DFS and DSS according to BRCA1/BRCA2 status and molecular phenotype [/fig_ref] in the multivariate analysis.
Response to neoadjuvant chemotherapy. Of the 263 (28%) patients who received neoadjuvant chemotherapy, the ER, PR and HER-2 status was available in 250 patients (95%). The pCR rate was significantly higher in BRCA1 (45%) compared to BRCA2 carriers (28%) and non-carriers (25%; p = 0.040; [fig_ref] Table 4: Pathologic complete response according to BRCA status and molecular subtype [/fig_ref]. Subgroup analysis by molecular subtype revealed that BRCA1 (54%) and BRCA2 carriers (57%) had significantly increased chemosensitivity compared to non-carriers (25%; p = 0.015) in the TNBC-subgroup only. In the HER-2 positive and the ER/PR positive/HER-2 negative subgroups, there was no difference between BRCA1/BRCA2 carriers and non-carriers regarding the pCR rate.
# Discussion
In the current study, we observed better disease-free survival of breast cancer patients who were selected for genetic screening, treated by chemotherapy and are BRCA carriers. Subgroup analysis revealed that the BRCA germline mutation is an independent prognostic factor associated with prolonged survival (both DFS and DSS) only for women with TNBC. For those who had ER/PR positive and/or HER-2 positive tumors (non-TNBC), BRCA mutations did not have any impact on outcome.
TNBC, mostly belonging to the basal-like subtype, share several molecular features of HGSOC including high levels of genomic instability and frequent TP53 mutations [bib_ref] Comprehensive molecular portraits of human breast tumours, Cancer Genome Atlas [/bib_ref] [bib_ref] High frequency of TP53 mutation in BRCA1 and sporadic basal-like carcinomas but..., Manie [/bib_ref]. The majority of HGSOC patients are diagnosed at advanced stages and receive platinum-based chemotherapy. BRCA carriers who developed HGSOC have increased survival compared to non-carriers [bib_ref] Association of BRCA1 and BRCA2 mutations with survival, chemotherapy sensitivity, and gene..., Yang [/bib_ref] [bib_ref] Association between BRCA1 and BRCA2 mutations and survival in women with invasive..., Bolton [/bib_ref]. This survival benefit has been linked to impaired DNA DSBs repair and consequently increased sensitivity to platinum [bib_ref] BRCAness revisited, Lord [/bib_ref]. For breast cancer patients, there are conflicting results regarding the prognosis and the predictive value of the BRCA germline status due to several issues: i) the phenotype of the tumor varies whether it is a BRCA1 (mainly TNBC) or a BRCA2 (mainly ER/PR positive) mutations; ii) adjuvant chemotherapy is not systematic and depends, among other characteristics, on tumor stage, grade, and molecular subtypes. Overall, it seems that BRCA1 carriers have poorer survival, probably due to the fact that they frequently develop TNBC, whereas BRCA2 germline mutation was not found to have a prognostic impact [bib_ref] Effects of BRCA1-and BRCA2-related mutations on ovarian and breast cancer survival: a..., Zhong [/bib_ref] [bib_ref] Effect of BRCA germline mutations on breast cancer prognosis: A systematic review..., Baretta [/bib_ref]. Whereas the prognostic value depends on tumor characteristics, the predictive value depends on the administered treatment. www.nature.com/scientificreports www.nature.com/scientificreports/ DNA interstrand crosslinks (ICLs) are among the most lethal lesions to DNA. They are generated by several chemotherapeutic drugs such as platinum, mitomycine and alkylating agents. Although these drugs are backbone therapy of multiple cancers, it is well after their introduction to the clinics that it was discovered that they act by inducing ICLs 28 . Cells defective in BRCA genes are highly sensitive to drugs that generate ICLs such as bifunctional alkylating agents and platinum [bib_ref] DNA interstrand crosslink repair and cancer, Deans [/bib_ref] [bib_ref] Targeting the DNA repair defect of BRCA tumours, Turner [/bib_ref] [bib_ref] Exploiting the DNA repair defect in BRCA mutant cells in the design..., Tutt [/bib_ref]. Another chemotherapeutic drugs that have biological background for efficacy in BRCA mutated tumors are topo-isomerase II inhibitors like anthracyclines [bib_ref] Chemotherapy for Patients with BRCA1 and BRCA2-Mutated Ovarian Cancer: Same or Different?, Tan [/bib_ref] [bib_ref] Phase II, open-label, randomized, multicenter study comparing the efficacy and safety of..., Kaye [/bib_ref]. Sensitivity to anthracylines and alkylating agents in BRCA carriers with breast cancer are emphasized by recent reports from INFORM and GeparOcto clinical trials [bib_ref] Association of Germline Variant Status With Therapy Response in High-risk Early-Stage Breast..., Pohl-Rescigno [/bib_ref] [bib_ref] TBCRC 031: Randomized Phase II Study of Neoadjuvant Cisplatin Versus Doxorubicin-Cyclophosphamide in..., Tung [/bib_ref].
We hypothesized that among breast cancer patients who received DNA damage agents BRCA carriers will be more chemosensitive and this could translate into survival benefit. A quarter of the patients in our cohort received neoadjuvant chemotherapy. We observed that pCR rates significantly differ according to BRCA1/BRCA2 status and molecular subtype. For TNBC, our data are consistent with previous reports showing increased pCR rate in BRCA1 9-12,33 and/or BRCA2 carriers [bib_ref] Germline Mutation Status, Pathological Complete Response, and Disease-Free Survival in Triple-Negative Breast..., Hahnen [/bib_ref] [bib_ref] BRCA1/2 Mutations and Bevacizumab in the Neoadjuvant Treatment of Breast Cancer: Response..., Fasching [/bib_ref] [bib_ref] Efficacy of anthracycline/taxane-based neo-adjuvant chemotherapy on triple-negative breast cancer in BRCA1/ BRCA2..., Bignon [/bib_ref]. However, less than half of BRCA carriers would develop TNBC 35 , 45% in our cohort, and few are known on chemosensitivity of BRCA2 carriers. We did not observe any impact of BRCA mutations on pCR in HER-2 positive or ER/PR-positive HER-2 negative tumors. ER/PR-positive tumors in BRCA2 carriers seemed resistant to chemotherapy with a response rate estimated to 7% only. Our observations should be interpreted cautiously given the limited number of patients in each subgroup and the highly selected population. Nevertheless, it suggests that chemosensitivity in BRCA carriers may dramatically vary with the molecular phenotype of the tumor 10,34,36 .
We observed a survival benefit in BRCA1/BRCA2 carriers who developed TNBC. There are contradicting results regarding the survival benefit of BRCA mutations in TNBC 6,9,12,16,17,33 . Plausible explanations are: i) we www.nature.com/scientificreports www.nature.com/scientificreports/ did not exclude BRCA2 carriers and they are rare compared to BRCA1, ii) our cohort of BRCA1/BRCA2 carriers who developed TNBC included more than 100 BRCA carriers. We did not observe any survival benefit in BRCA1/BRCA2 carriers with HER-2 positive or ER/PR-positive HER-2 negative breast cancers (non-TNBC). This result was unexpected and mirrors the response rates to neoadjuvant chemotherapy in the different subgroups. It suggests the existence of different types of breast tumors arising in BRCA carriers with distinct responses to DNA-damaging agents. Investigating the molecular mechanisms underlying these differences, such as mutational signatures [bib_ref] HRDetect is a predictor of BRCA1 and BRCA2 deficiency based on mutational..., Davies [/bib_ref] , somatic loss of the wild-type allele [bib_ref] BRCA locus-specific loss of heterozygosity in germline BRCA1 and BRCA2 carriers, Maxwell [/bib_ref] , BRCA genotype, the references doi: 10.1007/s10549-018-05127-2 and doi: 10.1158/1078-0432 recombination deficiency scores and/or infiltration by lymphocytes [bib_ref] Breast cancer genome and transcriptome integration implicates specific mutational signatures with immune..., Smid [/bib_ref] [bib_ref] Multifactorial analysis of differences between sporadic breast cancers and cancers involving BRCA1..., Lakhani [/bib_ref] are important questions that need to be addressed in the future.
Our results are consistent with the recently published POSH study, a large prospective cohort (>2,700) that addressed the prognostic value of BRCA mutations in young women (<40 years). The majority of participants (89%) and virtually all cases of TNBC (98%) received chemotherapy. The POSH study showed survival benefit only in BRCA carriers who had developed TNBC and this benefit was observed in the first two years following diagnosis [bib_ref] Germline BRCA mutation and outcome in young-onset breast cancer (POSH): a prospective..., Copson [/bib_ref]. The POSH study brings new insights into the prognostic value of BRCA mutations in the context of breast cancer in young women treated by chemotherapy. The Geparquinto trial consistently showed survival benefit from BRCA germline mutations in TNBC [bib_ref] BRCA1/2 Mutations and Bevacizumab in the Neoadjuvant Treatment of Breast Cancer: Response..., Fasching [/bib_ref].
Our study had several limitations. It is a retrospective study that included patients screened for BRCA1/ BRCA2 germline mutations. We recruited only women who were preselected based on their personal or family history that suggests a genetic predisposition. There might be a very specific additional risk factor profile for both environmental and genetic factors in these patients [bib_ref] Genome-wide association study in BRCA1 mutation carriers identifies novel loci associated with..., Couch [/bib_ref] [bib_ref] The role of genetic breast cancer susceptibility variants as prognostic factors, Fasching [/bib_ref]. In the French cohort we included all BRCA carriers and a subgroup of non-carriers who were randomly selected. This lead to a substantial enrichment of BRCA carriers among women with TNBC (45%), much higher than expected for unselected TNBC [bib_ref] Germline Mutation Status, Pathological Complete Response, and Disease-Free Survival in Triple-Negative Breast..., Hahnen [/bib_ref] [bib_ref] BRCA1/2 Mutations and Bevacizumab in the Neoadjuvant Treatment of Breast Cancer: Response..., Fasching [/bib_ref]. These biases are reflected by the young age of our cohort that does not represent the general population of breast cancer patients. There is a survival bias related to the time from cancer diagnosis to genetic testing. We excluded women who did not receive adjuvant chemotherapy and thus could not address the prognostic value of the BRCA status among this population. We probably missed a substantial proportion of BRCA carriers who did not undergo genetic screening due to the absence of personal or family history [bib_ref] Germline BRCA mutation and outcome in young-onset breast cancer (POSH): a prospective..., Copson [/bib_ref] [bib_ref] HRDetect is a predictor of BRCA1 and BRCA2 deficiency based on mutational..., Davies [/bib_ref]. Moreover, this study does not include a central review of pathology data. Nevertheless, BRCA carriers in our cohort had clinical and pathological characteristics consistent with previous reports [bib_ref] Pathology of breast and ovarian cancers among BRCA1 and BRCA2 mutation carriers:..., Mavaddat [/bib_ref] [bib_ref] Breast cancer prognosis in BRCA1 and BRCA2 mutation carriers: an International Prospective..., Goodwin [/bib_ref].
The strengths of our study are the following: we conducted a multicentric, international study with patients recruited in cancer comprehensive center, university hospitals, and private clinics. All patients underwent complete BRCA1 and BRCA2 gene sequencing, avoiding a selection bias in studies with founder mutations only [bib_ref] Clinical outcomes of breast cancer in carriers of BRCA1 and BRCA2 mutations, Rennert [/bib_ref]. We analyzed separately the impact of BRCA1 and BRCA2 mutations on survival and pCR and we did not focus on one molecular subtype or chemotherapy regimen or setting.
In summary, our study suggests that the prognostic value of BRCA1/BRCA2 germline mutations in breast cancer patients who were preselected for genetic screening and treated with neoadjuvant or adjuvant chemotherapy depends on the molecular subtype with a survival benefit only in women with TNBC. www.nature.com/scientificreports www.nature.com/scientificreports/ Swiss cohort). From Geneva, all women (BRCA carriers and non-carriers) who met the inclusion criteria were included. In order to reduce the number of non-BRCA carriers in the study cohort, all BRCA carriers and a subgroup of non-carriers diagnosed in Lyon (randomly selected) were included. A protocol with a standardized case report form was used for all data collection and submitted to the Geneva Commission cantonale d' ethique de la recherche (CCER 15-158). The study protocol was approved by the Geneva Commission cantonale d' ethique de la recherche and the local institutional review boards in both hospitals in France. Informed written consent was obtained from all patients in the French cohort, and all living patients in the Swiss cohort. The research was performed in accordance with relevant guidelines/regulations. Exclusion criteria were the absence of neoadjuvant or adjuvant chemotherapy, no genetic screening, no follow-up or metastatic disease at diagnosis. Data collection. Patient and treatment characteristics were collected from the medical records of patients treated at the Centre Leon Bérard, the Hospices Civils de Lyon, the Hôpitaux Universitaires de Genève and among 7 medical oncologists in private clinics in Geneva, Switzerland. We recorded date of birth, date of diagnosis, chemotherapy regimen, and timing (neoadjuvant or adjuvant). Chemotherapy agents were classified as anthracyclines, alkylating agents, taxanes, or platinum. Trastuzumab and hormonal therapy administration was recorded.
# Methods
Tumor characteristics were collected from pathological reports. This included histological subtype, grade, estrogen and progesterone receptors status (positivity was defined as nuclear staining of >1% by immunohistochemistry (IHC)), HER-2 status (defined as either 3+ by IHC or as assessed by gene amplification through fluorescence or chromogenic in situ hybridization). TNBC were defined as ER, PR and HER-2 negative tumors. Non-TNBC were defined as ER/PR and/or HER-2 positive tumors.
TNM staging was evaluated according to the timing of chemotherapy. If the patient received adjuvant chemotherapy, the pTNM was recorded. If the patient received neoadjuvant chemotherapy, the cTNM and yTNM were recorded. Axillary lymph nodes were considered positive if a pre-chemotherapy biopsy was positive or if there was at least one yN+ or the presence of a histological scar in the removed lymph nodes after neoadjuvant chemotherapy.
Genetic analysis. Women were referred to the genetic unit for complete BRCA1 and BRCA2 germline screening based on the presence of personal history of breast cancer presented at a young age, or the display of a particular tumor phenotype (TNBC) or association with ovarian cancer, or in the context of a positive family Statistical analyses. Based on a sample size of 600 non-carriers, 150 BRCA1 carriers and 100 BRCA2 carriers, a 80% 5-year DFS among non-carriers and a median follow-up of 6 years, the study had a 93% power to show an improvement of the 5-year DFS from 80% among non-carriers to 89% among the BRCA carriers (translating in a hazard ratio of 0.5) at a 2-sided alpha risk of 5%. DFS was calculated from the time of diagnosis until the date of first documented local, regional, or distal invasive recurrence or death from breast cancer, or to the time of last follow-up. DSS was defined as the time from diagnosis to death caused by breast cancer. Survival outcomes were estimated using the Kaplan-Meier product-limit method and compared by a long-rank test. Cox proportional-hazards (PH) models were fitted to determine the association of the BRCA germline status (with time to event outcomes before and after adjustment for significant patient and clinical characteristics. The proportional hazards hypothesis was assessed both graphically and statistically. Cox proportional-hazards models were used for all analyses given the absence of significant deviation from the PH hypothesis in all subgroups and for all reported outcome measures. The following prognostic variables were assessed in univariate analyses: BRCA status, age (≤ or> 35 years of age), lymph node status, SBR grade. Variables yielding p values less than 0.1 by univariate analysis were retained for the multivariate analysis. The proportional hazards assumption was assessed using scaled Schoenfeld residuals. Because of the high correlation between grade and lymph node involvement and in order to avoid colinearity, grade was not included in the multivariate model. P values of ≤ 0.05 were considered statistically significant. As a sensitivity analysis of the main outcomes, a landmark analysis was conducted to exclude patients with DFS or DSS of less than 12 months, in order to avoid any immortal time bias related to the time between the cancer diagnosis and the time of the genetic counseling/testing.
Pathological complete response (pCR) was defined as the absence of any invasive disease in the breast and in the ipsilateral axillary lymph nodes (ypT0/is ypN0) in accordance with the Union for International Cancer Control TNM system. Patient or tumor characteristics and chemotherapy regimens were compared according to the BRCA germline status using χ 2 tests or the Fisher's exact test for categorical variables, and non-parametric Kruskall-Wallis tests for continuous variables. All statistical analyses were carried out using the R software version 3.3.1 (http://www.r-project.org/).
## Data availability
All data analyzed during the study has been included in the manuscript (and its supplementary information files).
[fig] Figure 1: DFS and DSS according to BRCA1/BRCA2 status and molecular phenotype. (2020) 10:7073 | https://doi.org/10.1038/s41598-020-63759-1 [/fig]
[table] Table 4: Pathologic complete response according to BRCA status and molecular subtype. pCR: pathologic complete response. *HER-2 status missing in 12 cases. [/table]
[table] Table 3: Multivariate analysis of DFS and DSS in TNBC. NI: not included.history. Blood samples for germline DNA testing were obtained after a signed consent. All participants were screened for BRCA1 and BRCA2 mutations. BRCA1 and BRCA2 variants were classified as pathogenic according to the ENIGMA BRCA1/2 Gene Variant Classification Criteria (http://www.enigmaconsortium.org/). Women with variants of uncertain significance were considered as non-carriers. outcome measures. The primary objectives were to compare Disease-free survival (DFS) andDisease-specific survival (DSS) among breast cancer patients according to BRCA germline mutations. Secondary objectives were to compare i) DSS and DFS according to BRCA status in the TNBC and the non-TNBC population; ii) pCR according to molecular subtype (TNBC vs non-TNBC) and BRCA status in the subgroup of patients who received neoadjuvant chemotherapy. [/table]
|
The aim of the present study was to characterize the phenolic acids, flavonoids, and antioxidant properties of monofloral honey collected from five different districts in Bangladesh. A new high performance liquid chromatography (HPLC) equipped with a UV detector method was developed for the identification of the phenolic acids and flavonoids. A total of five different phenolic acids were identified, with the most abundant being caffeic acid, benzoic acid, gallic acid, followed by chlorogenic acid and transcinnamic acid. The flavonoids, kaempferol, and catechin were most abundant, followed by myricetin and naringenin. The mean moisture content, total sugar content, and color characteristics of the honey samples were 18.36 ± 0.95%, 67.40 ± 5.63 g/100 g, and 129.27 ± 34.66 mm Pfund, respectively. The mean total phenolic acids, total flavonoid content, and proline content were 199.20 ± 135.23, 46.73 ± 34.16, and 556.40 ± 376.86 mg/kg, respectively, while the mean FRAP values and DPPH radical scavenging activity were 327.30 ± 231.87 M Fe (II)/100 g and 36.95 ± 20.53%, respectively. Among the different types of honey, kalijira exhibited the highest phenolics and antioxidant properties. Overall, our study confirms that all the investigated honey samples are good sources of phenolic acids and flavonoids with good antioxidant properties.
# Introduction
Honey is a natural product consisting of a highly concentrated solution of a complex mixture of sugar and minute quantities of other constituents, such as minerals, proteins, vitamins, organic acids, flavonoids, phenolic acids, enzymes, and volatile compounds [bib_ref] Antioxidant activities and total phenolics of different types of honey, Al-Mamary [/bib_ref] [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref]. The quantity of these different compounds varies greatly depending on the floral and geographical origin of the honey. Additionally, the composition of honey is influenced by processing, handling, and storage time [bib_ref] Evaluation of the phenolic content, antioxidant activity and colour of Slovenian honey, Bertoncelj [/bib_ref] [bib_ref] Identification and quantification of antioxidant components of honeys from various floral sources, Gheldof [/bib_ref].
The components in honey reported to be responsible for its antioxidant effects are flavonoids, phenolic acids, ascorbic acid, catalase, peroxidase, carotenoids, and the products of Maillard reactions [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref] [bib_ref] Identification and quantification of antioxidant components of honeys from various floral sources, Gheldof [/bib_ref]. However, the amount and type of these antioxidants are largely dependent on the floral source or honey variety and a correlation between antioxidant activity and total phenolic content has been established [bib_ref] Identification and quantification of antioxidant components of honeys from various floral sources, Gheldof [/bib_ref].
Bangladesh is a country with a subtropical monsoon climate and experiences wide seasonal variations in rainfall, moderately warm temperatures, and high humidity. Honey has been used traditionally over the years by the people of Bangladesh as food and as a traditional medicine in the treatment of several diseases. Although honey is widely consumed by locals, very few data are available to support the medicinal claims of different types of honey samples from Bangladesh. Several types of honey, such as mustard (Brassica nigra), Kalijira (Nigella sativa), Padma flower (Nelumbo nucifera), sesame (Sesamum indicum), drumstick (Moringa oleifera), blackberry (Syzygium cumini), and lychee (Litchi chinensis), are available in Bangladesh. Although honey is widely consumed in Bangladesh, few data are available on the quality of commonly consumed honey. Investigations of honey samples collected from different floral sources and geographic locations are necessary to provide local data. Furthermore, the data available for honey reported from other countries are not applicable to Bangladesh because honey varies in antioxidant capacity, physicochemical properties, and composition based on its floral sources. In this study, we aimed to investigate different types of honey samples collected from different regions in Bangladesh and to identify the different phenolic acids, flavonoids, and antioxidant properties of some monofloral honey samples from Bangladesh. To the best of our knowledge, this is the first study to extensively investigate the different types of antioxidants present in various types of honey samples from Bangladesh.
# Materials and methods
## Honey samples.
A total of ten monofloral honey ( = 10) samples (BH-1 to BH-10) were collected from eight different locations in five different districts (Tangail, Jamalpur, Khulna, Madaripur, and Munshigonj) in Bangladesh [fig_ref] Figure 1: Sources of honey. [/fig_ref]. The details of the honey, including the honey's local and scientific names, are described in [fig_ref] Table 1: Floral type and source of the investigated Bangladeshi monofloral honey [/fig_ref]. All honey collections were performed between January and December 2012. The samples were refrigerated (4-5 ∘ C) in airtight plastic containers until further analysis. All analyses were conducted in triplicate.
## Chemicals and reagents.
The phenolic acids (gallic, syringic, caffeic, vanillic, benzoic, and trans-cinnamic acids) and flavonoids (catechin, naringenin, luteolin, hesperetin, kaempferol, apigenin, and naringin), 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,4,6-tris(1-pyridyl)-1,3,5-triazine (TPTZ), Folin-Ciocalteu's reagent, and gallic acid were purchased from Sigma-Aldrich (St. Louis, MO, USA). Sodium carbonate (Na 2 CO 3 ), aluminum chloride (AlCl 3 ), sodium nitrite (NaNO 2 ), and sodium hydroxide (NaOH) were purchased from Merck (Darmstadt, Germany). All chemicals were of analytical grade.
# Physical analysis
2.3.1. Moisture Content. The moisture content was determined by using a refractometric method. In general, the refractive index increases with an increase in the solid content of the sample. The refractive indices of honey samples were measured at ambient temperature using an Atago handheld refractometer (KRUSS, HRH30, Hamburg, Germany). The measurements were further corrected for the standard temperature of 20 ∘ C by adding a correction factor of 0.00023/ ∘ C. The moisture content was measured in triplicate and the percentage of moisture content that corresponds to the corrected refractive index was calculated using Wedmore's table.
## Total sugar content.
Honey was suspended in Milli-Q water to make a 25% (w/v) solution. The total sugar content of each honey sample was then determined using a refractometric method (Atago handheld refractometer, ATAGO, N-1 , Tokyo, Japan). The percentage of sucrose content was measured per g/mL of honey.
## Honey color
Analysis. The color intensity of honey samples was measured according to the Pfund classifier. Briefly, homogeneous honey samples devoid of air bubbles were transferred into a cuvette with a 10 mm light path until the cuvette was approximately half full. The cuvette was inserted into a color photometer (HI 96785, Hanna Instrument, Cluj County, Romania). Color grades were expressed in millimeter (mm) Pfund grades when compared to an analytical-grade glycerol standard. Measurements were performed in triplicate for each sample using the approved color standards of the United States Department of Agriculture (USDA). [bib_ref] Standardization of antioxidant properties of honey by a combination of spectrophotometric/fluorimetric assays..., Beretta [/bib_ref]. Briefly, honey samples were diluted to 50% (w/v) with warm (45-50 ∘ C) Milli-Q water and the resulting solution was filtered using a 0.45 m filter to remove large particles. The absorbance was measured at 450 and 720 nm using a spectrophotometer and the difference in absorbance was expressed as mAU.
## Analysis of antioxidant properties
## Determination of total phenolic
Compounds. The concentration of phenolic compounds in honey samples was estimated using a modified spectrophotometric Folin-Ciocalteu method [bib_ref] Analysis of total phenols and other oxidation substrates and antioxidants by means..., Singleton [/bib_ref]. Briefly, 1 mL of honey extract was mixed with 1 mL of Folin and Ciocalteu's phenol reagent. After 3 min, 1 mL of Na 2 CO 3 (10%) solution was added to the mixture and adjusted to 10 mL with distilled water. The reaction was kept in the dark for 90 min, after which the absorbance was read at 725 nm using a T 60 UV/VIS spectrophotometer (PG Instruments Ltd., UK). Gallic acid was used to calculate a standard curve (20, 40, 60, 80, and 100 g/mL; 2 = 0.996). The concentration of phenolic compounds was measured in triplicate. The results were reported as mean ± standard deviation and expressed as mg of gallic acid equivalents (GAEs) per kg of honey.
## Determination of total flavonoid content.
The total flavonoid content in each honey sample was measured using the colorimetric assay developed by Zhishen et al. [bib_ref] The determination of flavonoid contents in mulberry and their scavenging effects on..., Zhishen [/bib_ref]. Honey extract (1 mL) was mixed with 4 mL of distilled water. At the baseline, 0.3 mL of NaNO 2 (5%, w/v) was added. After five min, 0.3 mL of AlCl 3 (10% w/v) was added, followed by the addition of 2 mL of NaOH (1 M) 6 min later. The volume was then increased to 10 mL by the addition of 2.4 mL distilled water. The mixture was vigorously shaken to ensure adequate mixing and the absorbance was read at 510 nm. A calibration curve was created using a standard solution of catechin [bib_ref] Physicochemical and antioxidant properties of algerian honey, Khalil [/bib_ref] , 60, 80, and 100 g/mL; 2 = 0.998). The results were expressed as mg catechin equivalents (CEQ) per kg of honey.
## Ferric reducing/antioxidant power assay (frap assay).
The FRAP assay was performed according to a modified method described by Benzie and Strain [bib_ref] Ferric reducing/antioxidant power assay: direct measure of total antioxidant activity of biological..., Benzie [/bib_ref]. Briefly, 200 L of properly diluted honey (0.1 g/mL) was mixed with 1.5 mL of FRAP reagent. The reaction mixture was then incubated at 37 ∘ C for 4 min and its absorbance was read at 593 nm against a blank that was prepared with distilled water. Fresh FRAP reagent was prepared by mixing 10 volumes of 300 mM/L acetate buffer (pH 3.
## Dpph free
Radical-Scavenging Activity. The antioxidant properties of each honey sample were also investigated by determining the free radical-scavenging activity of the DPPH radical based on the method proposed by Ferreira et al. [bib_ref] Antioxidant activity of Portuguese honey samples: different contributions of the entire honey..., Ferreira [/bib_ref]. Briefly, honey extract (0.5 mL) was mixed with 2.7 mL of methanolic solution containing DPPH radicals (0.024 mg/mL). The mixture was vigorously shaken and left to stand for 15 min in the dark (until the absorbance stabilized). The reduction of the DPPH radical was determined by measuring the absorbance of the mixture at 517 nm. Butylated hydroxytoluene (BHT) was used as a reference. The radical-scavenging activity (RSA) was calculated as the percentage of DPPH discoloration using the following equation: % RSA = ([ DPPH − ]/ DPPH ) × 100, where is the absorbance of the solution when the sample extract is added at a particular level and DPPH is the absorbance of the DPPH solution.
## Proline content.
The proline content in the honey samples was measured using a method established by the International Honey Commission (IHC). Briefly, approximately 5 g of honey was transferred to a beaker and dissolved in 50 mL water. The solution was quantitatively transferred to a 100 mL volumetric flask before further dilution to 100 mL with distilled water. After that, approximately 0.5 mL of the sample solution was transferred to a tube, while 0.5 mL of water (blank test) was transferred to a second tube and 0.5 mL of proline standard solution was dispensed into three other tubes. To each tube, approximately 1 mL of formic acid and 1 mL of ninhydrin solution were added. The tubes were capped carefully and shaken vigorously for 15 min. The tubes were then placed in a boiling water bath for 15 min and immersed below the level of the solution. The tubes were further transferred to another water bath and incubated at 70 ∘ C for 10 min. Approximately 5 mL of the 2-propanol water solution was added to each tube followed by immediate capping. The tubes were left to cool for approximately 45 min after removal from the 70 ∘ C water bath and the absorbance values were measured at 510 nm (near the maximum wavelength).
## Hplc analysis of phenolic acids and flavonoids
## Extraction of phenolic compounds.
A modified solidphase extraction (SPE) procedure was developed to extract the phenolic compounds present in honey. Briefly, honey (2 g) was dissolved in 10 mL of acidified deionized water (pH 2) that was pH adjusted using orthophosphoric acid (85%) and the solution was run through Bond Elut C18 cartridges (3 mL × 500 mg) (Agilent Technologies, Santa Clara, CA, USA). The cartridges were preconditioned by sequentially passing 3 mL of methanol followed by the addition of acidified water (pH 2) dropwise. An aqueous honey solution (5 mL) was then applied to the preconditioned cartridges drop by drop to ensure efficient adsorption of the investigated compounds to the bonded phase. The adsorbed phenolic compounds were then eluted from the cartridges with 2 mL of 90% (v/v) methanol/water solution dropwise. The eluates were collected and evaporated to dryness under a gentle flow of nitrogen gas. Finally, the extract was reconstituted with 1 mL of methanol before HPLC analysis.
# Hplc analysis.
A new HPLC method was developed for the detection of phenolic acids and flavonoids by employing an HPLC system (Waters 2695, Milford, MA, USA) equipped with a photodiode array detector (Waters 2996). The HPLC column was a Merck Purospher Star, RP-18e (150 × 4.6 mm, 5 m), fitted with a guard cartridge that had been packed with a similar type of stationary phase (Merck). A linear gradient flow was employed at a flow rate of 0.5 mL/min throughout and the total analytical time was approximately 40 min.
The binary mobile phase consisted of solvent A (ultrapure water with 0.1% phosphoric acid) and solvent B (pure methanol with 0.1% phosphoric acid). Elution from the column was achieved with the following gradients: 0 to 20 min of solvent B, increasing from 10% to 85%; 20 to 25 min of solvent B at 85% throughout; 25 to 26 min of solvent B from 85% decreasing to 10%; and a final composition at 10% that was kept constant to 40 min. The detection wavelength was set between 200 and 450 nm, with specific monitoring conducted at 220 nm. Identification of the phenolic and flavonoid compounds was performed by comparing the retention times of the analytes with reference standards.
# Statistical analysis.
Assays were performed in triplicate and the results were expressed as the mean values with standard deviations (SD). The significant differences represented by letters were obtained by a one-way analysis of variance (ANOVA) followed by Tukey's honestly significant difference (HSD) post hoc test ( < 0.05). Correlations were established using Pearson's correlation coefficient ( ) in bivariate linear correlations ( < 0.01). These correlations were calculated using Microsoft office Excel 2007 and SPSS version [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref].0 (IBM corporation, New York, USA).
# Results and discussion
[formula] 3.1. Moisture Content. [/formula]
Moisture is a physicochemical parameter of honey associated with the climatic conditions of the location from which the honey was gathered as well as the degree of honey maturity [bib_ref] Composition of honey from Córdoba (Argentina): assessment of North/South provenance by chemometrics, Baroni [/bib_ref]. It is one of the key factors that determine the quality of honey. In the present investigation, the moisture content was between 16.33 and 19.53% [fig_ref] Figure 2: Physical parameters such as [/fig_ref] ), which was within the limit of ≤20% set by the international regulations for honey quality [bib_ref] Revised codex standard for honey, Alimentarius [/bib_ref]. The moisture content present in honey samples is important because it contributes to the honey's ability to resist fermentation and crystallization during storage [bib_ref] The correlation between water activity and % moisture in honey: fundamental aspects..., Chirife [/bib_ref] [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref].
Significant variations in moisture content were observed in the investigated monofloral honey. The mean moisture content was 18.36 ± 0.95%, with the lowest moisture content (16.33%) exhibited by sample BH-8, which was collected in June 2012, while sample BH-9 exhibited the highest moisture content (19.53%). Low moisture content in honey can confer a protective effect against microbial attack, especially during long term storage [bib_ref] The correlation between water activity and % moisture in honey: fundamental aspects..., Chirife [/bib_ref] [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref]. On the other hand, high moisture content can promote honey fermentation during storage as a result of the activity of osmotolerant yeasts and the consequent formation of ethyl alcohol and carbon dioxide [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref]. The lower moisture content exhibited by most of the investigated honey samples ensures the better quality of these honey samples, which allows them to be stored for a longer duration.
The moisture contents of the analyzed samples were consistent with previously reported values of honey samples from Bangladesh of 12.79% to 22.32% [bib_ref] Physicochemical and antioxidant properties of Bangladeshi honeys stored for more than one..., Islam [/bib_ref]. Moreover, the moisture contents for the investigated honey samples were similar to those from countries of similar climates, including honey from India (17.20 to 21.60%) [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref] and Malaysia (14.86-17.53%) [bib_ref] Physicochemical and antioxidant properties of Malaysian honeys produced by Apis cerana, Apis..., Moniruzzaman [/bib_ref].
## Total sugar content.
The total sugar content of the investigated honey samples ranged from 55.67 to 73.67%, with a mean value of 67.40 ± 5.63% [fig_ref] Figure 2: Physical parameters such as [/fig_ref]. According to the European Commission directive, the total glucose and fructose content of honey should exceed 60 g per 100 g of honey for natural honey. All of the investigated honey samples contained total sugar contents higher than the recommended level, except for sample BH-10 (55.67%). The lower total sugar content of sample BH-10 can be attributed to the processing or storage and cannot be attributed to the honey source because sample BH-1, which contained the highest total sugar content (73.67%), was from a similar source. Sugars generally constitute the main components of honey, regardless of type, with reducing sugars (mainly fructose and glucose) making up the majority. In a previous report, the total sugar content of Bangladeshi honey samples was low, ranging from 42.80 to 60.67% [bib_ref] Physicochemical and antioxidant properties of Bangladeshi honeys stored for more than one..., Islam [/bib_ref]. Although the honey came from the same country, the variation observed could have been due to the different floral sources. However, the sugar content of the investigated honey samples is reported to be generally lower than that reported for honey from Algeria (62.80 to 70.00%), which tends to be sweeter [bib_ref] Physicochemical and antioxidant properties of algerian honey, Khalil [/bib_ref].
## Color characteristics.
Honey color based on the USDAapproved color standardsis one of the primary characteristics of honey classification. Honey color differs naturally, ranging from light yellow to amber, dark amber, and black (in extreme cases) and sometimes green or red hues may even occur [bib_ref] Physicochemical and antioxidant properties of Malaysian honeys produced by Apis cerana, Apis..., Moniruzzaman [/bib_ref]. The color of untreated honey depends on its botanical origins. For this reason, color is very important for the classification of monofloral honey for commercial use.
The higher the Pfund value is, the darker the honey color should be. In the present study, the honey color ranged from 70.33 to 150.00 mm Pfund . The majority of the investigated honey samples were dark amber in color, with a mean Pfund value of 150.00, indicating their better qualities, as honey color is reported to be dependent not only on various components, such as ash and potential alkalinity, but also on the concentrations of antioxidants, such as phenolic acids and flavonoids. Our results are similar to those reported for some honey samples from Malaysia (38.33-150.00) [bib_ref] Physicochemical and antioxidant properties of Malaysian honeys produced by Apis cerana, Apis..., Moniruzzaman [/bib_ref] and Algeria (107.00-150.00) [bib_ref] Physicochemical and antioxidant properties of algerian honey, Khalil [/bib_ref].
## Color
Intensity. The color intensity of honey is represented by the ABS 450 , which also indicates the presence of pigments such as carotenoids and flavonoids, which are also known to contribute to antioxidant properties [bib_ref] Physicochemical and antioxidant properties of Malaysian honeys produced by Apis cerana, Apis..., Moniruzzaman [/bib_ref] [bib_ref] Physicochemical and antioxidant properties of algerian honey, Khalil [/bib_ref]. In the present study, the honey's ABS 450 values ranged between 117 and 1174 mAU [fig_ref] Figure 2: Physical parameters such as [/fig_ref]. Sample BH-3, which was among the highest Pfund values, also showed correspondingly the highest color intensity (1174.00 ± 3.61 mAU), indicating its better antioxidant properties. High color intensity was also observed in two other honey samples (BH-10 and BH-9 at 841.67 and 564.33 mAU, resp.), which correspondingly had high antioxidant potential ,as also reported by Comparatively, the ABS 450 values were reported to be between 724 and 1188 mAU in honey samples from Algeria [bib_ref] Physicochemical and antioxidant properties of algerian honey, Khalil [/bib_ref] , between 25 and 3413 mAU in honey samples from Italy [bib_ref] Standardization of antioxidant properties of honey by a combination of spectrophotometric/fluorimetric assays..., Beretta [/bib_ref] , between 70 and 495 mAU in honey samples from Slovenia [bib_ref] Evaluation of the phenolic content, antioxidant activity and colour of Slovenian honey, Bertoncelj [/bib_ref] , between 254 and 2034 mAU in honey samples from Bangladesh [bib_ref] Physicochemical and antioxidant properties of Bangladeshi honeys stored for more than one..., Islam [/bib_ref] , and between 524 and 1678 mAU in honey samples from India [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref]. The comparatively high ABS 450 values in the investigated honey samples indicate their high antioxidant properties and purity.
## Total phenolic content.
The concentration and type of phenolic substances depend on the floral origin of the honey and are mainly responsible for its biological activities [bib_ref] Antioxidant activities and total phenolics of different types of honey, Al-Mamary [/bib_ref]. The total phenolic content (mg/kg of honey) was found to vary significantly among the various honey samples from Bangladesh , which may be due to their different botanical and regional origins. The mean total phenolic content of the studied honey samples was 199.20 mg/kg, with the highest phenolic content recorded by sample BH-3 at 465.68 ± 3.75 mg/kg, indicating its superior antioxidant potential. kalijira honey (BH-3) which is only available in the winter season in Bangladesh is significantly different from the samples analysed. Two previous studies [bib_ref] Antioxidant activity and phenolic content of phenolic rich fractions obtained from black..., Mariod [/bib_ref] [bib_ref] Polyphenolic profile and antioxidant activities of Nigella sativa seed extracts in vitro..., Meziti [/bib_ref] reported that black cumin (Nigella sativa) is a good source of phenolic compounds and exhibits high antioxidant capacity. It is plausible that the high phenolic content in the plant parts of the black cumin (Nigella sativa) tree is transferred into the nectar collected by the bees and introduced to the honey, contributing to the high phenolic contents observed in this type of honey.
The mean phenolic content of the investigated honey samples is higher than that of lime honey (83.7 mg/kg) [bib_ref] Evaluation of the phenolic content, antioxidant activity and colour of Slovenian honey, Bertoncelj [/bib_ref] , acacia honey from Slovenia (44.8 mg/kg) [bib_ref] Evaluation of the phenolic content, antioxidant activity and colour of Slovenian honey, Bertoncelj [/bib_ref] , and honey samples from Burkina Faso (74.38 mg/kg) [bib_ref] Determination of the total phenolic, flavonoid and proline contents in Burkina Fasan..., Meda [/bib_ref] , indicating its superior antioxidant properties. In a previous study, the phenolic contents of honey samples from Bangladesh that were stored for more than one year were slightly higher (152.4 to 688.5 mg/kg) [bib_ref] Physicochemical and antioxidant properties of Bangladeshi honeys stored for more than one..., Islam [/bib_ref]. It has been reported that the botanical [bib_ref] Evaluation of the phenolic content, antioxidant activity and colour of Slovenian honey, Bertoncelj [/bib_ref] [bib_ref] Standardization of antioxidant properties of honey by a combination of spectrophotometric/fluorimetric assays..., Beretta [/bib_ref] and geographical region from which the honey is collected not only affects its phenolic and flavonoid concentrations but also affects its pollen distribution and, ultimately, its antioxidant activities.
## Total flavonoid content.
Flavonoids are low molecular weight phenolic compounds responsible for the aroma and the antioxidant potential of honey. The total flavonoid content in the investigated honey samples ranged from 11.46 to 116.67 mg catechin /kg . As with the phenolic content, the highest flavonoid concentration was also shown by sample BH-3. The results are in agreement with those of several previous studies, in which it was found that honey samples with higher polyphenol content will also yield high flavonoid levels [bib_ref] Physicochemical and antioxidant properties of Malaysian honeys produced by Apis cerana, Apis..., Moniruzzaman [/bib_ref] [bib_ref] Physicochemical and antioxidant properties of algerian honey, Khalil [/bib_ref]. On average, the mean flavonoid content (46.73 mg/kg) of the investigated honey samples from Bangladesh was higher than that of honey samples from Slovenia (20.57 mg/kg) [bib_ref] Determination of the total phenolic, flavonoid and proline contents in Burkina Fasan..., Meda [/bib_ref] , indicating the superior antioxidant properties of these honey samples.
## Determination of total antioxidant content by frap
Assay. The FRAP assay is used to measure the total antioxidant content of honey. It directly estimates the presence of either antioxidants or reductants, depending on the ability of the analyte to reduce the Fe 3+ /Fe 2+ redox couple [bib_ref] Standardization of antioxidant properties of honey by a combination of spectrophotometric/fluorimetric assays..., Beretta [/bib_ref]. The FRAP values for the investigated honey samples from Bangladesh ranged from 116.00 to 786.22 M Fe (II)/100 g of honey. There were significant differences among the FRAP values for the different types of honey
## Proline content.
Proline is one of the most abundant amino acids in honey and is therefore usually selected as the standard for quantifying the amino acid content. Moreover, the proline content is an indicator for honey and can also reveal adulteration, as indicated by a proline level of less than 183 mg/kg [bib_ref] Determination of the total phenolic, flavonoid and proline contents in Burkina Fasan..., Meda [/bib_ref]. Proline content in honey is a criterion of honey ripeness and robustness. In our study, all of the investigated honey samples contained high amounts of proline, which ranged from 237.51 to 1419.33 mg/kg . Similar to the phenolic acid and flavonoid contents, honey sample BH-3 again contained the highest proline content (1419 mg/kg), indicating its superior antioxidant potential.
Generally, the proline contents of the investigated samples were higher than those previously reported from Bangladesh (106-681 mg/kg) [bib_ref] Physicochemical and antioxidant properties of Bangladeshi honeys stored for more than one..., Islam [/bib_ref] , India (133-674 mg/kg) [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref] , and Malaysia 392.85 mg/kg [bib_ref] Physicochemical and antioxidant properties of Malaysian honeys produced by Apis cerana, Apis..., Moniruzzaman [/bib_ref].
## Dpph free radical-scavenging activity.
In evaluating the radical-scavenging potential of a sample including honey, the DPPH assay is frequently used. Usually, a high DPPH scavenging activity reflects high levels of antioxidant potential. The mean DPPH radical-scavenging activity of the investigated honey samples was 36.95%. Sample BH-3 again exhibited the highest DPPH radical-scavenging activity (76.68%), which could be attributed to its higher phenolic acid and flavonoid content , as it has been reported that the antioxidant potential of honey is directly proportional to the amount of phenolic acids and flavonoids present [bib_ref] Standardization of antioxidant properties of honey by a combination of spectrophotometric/fluorimetric assays..., Beretta [/bib_ref]. Overall, the DPPH scavenging and antioxidant potentials of the honey samples from Bangladesh were higher than that previously reported for some Malaysian honey samples [bib_ref] Physicochemical and antioxidant properties of Malaysian honeys produced by Apis cerana, Apis..., Moniruzzaman [/bib_ref] , Indian honey samples [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref] , and Algerian honey samples [bib_ref] Physicochemical and antioxidant properties of algerian honey, Khalil [/bib_ref] , again indicating their good quality.
## Identification and determination of phenolic compounds by hplc.
A modified solid-phase extraction (SPE) sample preparation method was developed for the extraction of phenolic compounds in honey based on the method published by Khalil et al. [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref]. A total of thirteen different phenolic compounds consisting of seven flavonoids (catechin, naringin, myricetin, naringenin, hesperetin, kaempferol, and apigenin) and six phenolic acids (gallic acid, chlorogenic acid, caffeic acid, coniferic acid, benzoic acid, and trans-cinnamic acid) were investigated [fig_ref] Figure 7: Typical chromatograms for [/fig_ref]. From this number, five phenolic acids (gallic acid, chlorogenic acid, caffeic acid, benzoic acid, and trans-cinnamic acid) and four flavonoids (catechin, myricetin, naringenin, and kaempferol) were detected in the honey samples from Bangladesh. The differences in the presence of the phenolic compounds in the investigated monofloral honey from Bangladesh may be due to their diverse botanical and regional sources.
Among the samples that contained between five and six phenolic acids were samples BH-3, BH-5, BH-8, and BH-9, corresponding with their high color intensity, total phenolic acids, total flavonoids, FRAP assay, and DPPH scavenging activities, as previously shown. However, based on the findings by Aljadi and Yusoff [bib_ref] Isolation and identification of phenolic acids in Malaysian honey with antibacterial properties, Aljadi [/bib_ref] , the antioxidant potential of honey as well as of propolis is mainly contributed by phenolic acids, as was also shown in our correlation analysis in the later part of the study.
Caffeic acid and benzoic acid were the most abundant phenolic compounds (70%) among all the phenolic acids and were detected in seven of the samples. This was followed by kaempferol (detected in five honey samples) and gallic acid (detected in four honey samples). Catechin, chlorogenic acid, myricetin, and naringenin were each detected in three of the honey samples, while apigenin was detected in a single sample. Coniferic acid, naringenin, and hesperetin were not present in any of the investigated honey samples. In a previous study, hesperetin and apigenin were detected in honey samples from Portugal [bib_ref] Analysis of honey phenolic acids by HPLC, its application to honey botanical..., Andrade [/bib_ref] , while apigenin was identified previously in a type of honey from Malaysia named gelam honey [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref]. In addition to apigenin, honey samples from Malaysia have also been reported to contain catechin, gallic acid, caffeic acid, syringic acid, benzoic acid, naringenin, trans-cinnamic acid, and kaempferol apigenin [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref]. In another study, gelam and coconut honey samples from Malaysia were found to contain gallic, caffeic, benzoic, ferulic, and cinnamic acids [bib_ref] Isolation and identification of phenolic acids in Malaysian honey with antibacterial properties, Aljadi [/bib_ref]. Honey samples from Australia contained gallic acid, caffeic acid, chlorogenic acid, myricetin, kaempferol, coumaric acid, ferulic acid, and quercetin [bib_ref] Phenolic acids in Australian Melaleuca, Guioa, Lophostemon, Banksia and Helianthus honeys and..., Yaoa [/bib_ref]. Honey samples from Portugal contained coumaric acid, ferulic acid, quercetin, vanillic acid, rosmarinic acid, and kaempferol [bib_ref] Analysis of honey phenolic acids by HPLC, its application to honey botanical..., Andrade [/bib_ref] [bib_ref] Antioxidant and antimicrobial effects of phenolic compounds extracts of Northeast Portugal honey, Estevinho [/bib_ref]. The variations observed may be due to the different botanical sources of honey from Bangladesh compared with honey from other parts of the world.
Kaempferol, gallic acid, and catechin are well known for their antioxidant properties, as described in previous studies [bib_ref] Isolation and identification of phenolic acids in Malaysian honey with antibacterial properties, Aljadi [/bib_ref] [bib_ref] Antioxidant and antimicrobial effects of phenolic compounds extracts of Northeast Portugal honey, Estevinho [/bib_ref]. The phenolic compounds present in the investigated honey samples from Bangladesh possess several medicinal properties, such as antioxidant, antibacterial, and antimicrobial properties. In particular, benzoic acid, chlorogenic acid, and caffeic acid were found to exhibit antibacterial properties [bib_ref] Isolation and identification of phenolic acids in Malaysian honey with antibacterial properties, Aljadi [/bib_ref] [bib_ref] Antioxidant and antimicrobial effects of phenolic compounds extracts of Northeast Portugal honey, Estevinho [/bib_ref].
The most abundant phenolic acids were caffeic acid (0.00-2.66 mg/kg), benzoic acid (0.00-2.31 mg/kg), gallic acid (0.00-0.61 mg/kg) and chlorogenic acid (0.00-1.35 mg/kg), followed by trans-cinnamic acid (0.00-0.85 mg/kg). The gallic acid contents in the investigated Bangladeshi honey are similar to those of honey samples reported from Malaysia (0.43 mg/kg) [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref] but lower than those of Australian honey (1.58 mg/100 g) [bib_ref] Phenolic acids in Australian Melaleuca, Guioa, Lophostemon, Banksia and Helianthus honeys and..., Yaoa [/bib_ref] , while chlorogenic acid was detected in Polish honey at 0.098 to 3.342 mg/100 g [bib_ref] Phenolic compounds and abscisic acid as potential markers for the floral origin..., Jasicka-Misiak [/bib_ref]. Similar to our findings, trans-cinnamic acid was reported at 0.01 to 0.50 mg/kg in Malaysian honey samples [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref]. Benzoic acid was found at 0.20-11.33 mg/kg in different honey samples from Malaysia [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref]. On the other hand, caffeic acid was found at 0.001 mg/kg in Malaysian Borneo tropical honey [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref] , 1.08 mg/100 g in Australian honey [bib_ref] Phenolic acids in Australian Melaleuca, Guioa, Lophostemon, Banksia and Helianthus honeys and..., Yaoa [/bib_ref] , and 0.021-0.101 mg/100 g in Polish honey [bib_ref] Phenolic compounds and abscisic acid as potential markers for the floral origin..., Jasicka-Misiak [/bib_ref].
For flavonoids, kaempferol (0.00-3.01 mg/kg) and catechin (0.00-1.99 mg/kg) were most abundant, followed by myricetin (0.00-0.49 mg/kg) and naringenin (0.00-0.38 mg/kg). Kaempferol was detected at 0.02 to 0.81 mg/kg in Malaysian honey [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref] and 0.01 to 0.099 mg/100 g in Polish honey [bib_ref] Phenolic compounds and abscisic acid as potential markers for the floral origin..., Jasicka-Misiak [/bib_ref].
The majority of phenolic compounds are small molecular weight compounds that have the tendency to elute faster from the column, with retention times varying between 8 and 24 min. When some of the compounds coeluted during the analysis, gradient elution was employed to produce better separated peaks and facilitate identification. In addition, there were some unknown peaks having similar spectra to those of flavonoids and phenolic acids. Unfortunately, they could not be fully identified due to the lack of standards for HPLC for these compounds.
Previous reports on phenolics compound concentrations in honey that utilized spectrophotometry methods [bib_ref] Phenolic acid composition and antioxidant properties of Malaysian honeys, Khalil [/bib_ref] [bib_ref] Approach to the content of total extractable phenolic compounds from different food..., Escarpa [/bib_ref] typically reported higher levels of phenolic compounds compared with those determined by HPLC methods. This could be due to the interference of nonphenolic materials present in the extracts, which may interfere with spectrophotometric analysis by causing higher readings, as suggested by Escarpa and González [bib_ref] Approach to the content of total extractable phenolic compounds from different food..., Escarpa [/bib_ref]. Overall, the phenolic compounds detected in the honey samples from Bangladesh have high antioxidant properties and may show promising pharmacological effects in the treatment of chronic diseases, which should be further confirmed in clinical trials.
3.11. Correlation Analysis. Several strong correlations were observed amongst some of the biochemical and antioxidant parameters. A strong correlation was established between the color intensities of the honey samples and their antioxidant parameters: phenolic acids, flavonoids, DPPH, and FRAP values at 0.943, 0.926, 0.838, and 0.894. The color intensity of the honey also increased with higher phenolic and flavonoid content in the honey. For example, BH-3 honey, which had the highest color intensity, also showed the highest phenolic content, suggesting that honey color pigments conferred by the phenolic compounds may play a role in the observed antioxidant activities of honey samples.
Another strong correlation was established between the ABS 450 , DPPH, and FRAP values, again indicating the involvement of pigments that ultimately confer antioxidant potential to honey. In a previous study conducted by , a strong correlation (r = 0.850) between the ABS 450 and FRAP values was established in Slovenian honey. The correlation between the ABS 450 and FRAP values was also high (r = 0.83) in Indian honey [bib_ref] Physical, biochemical and antioxidant properties of some Indian honeys, Saxena [/bib_ref] , indicating that ABS 450 , DPPH, and FRAP values are good predictors of antioxidant properties in honey. Thus, the higher correlations established in our study (ABS 450 and DPPH, r = 0.838; ABS 450 and FRAP, 0.894) suggest that the investigated monofloral honey samples from Bangladesh have a stronger antioxidant capacity when compared to Indian and Slovenian honey.
A positive significant linear correlation was also observed for the phenolic and flavonoid contents with the DPPH ( = 0.915) and FRAP values ( = 0.876), respectively, which indicates the involvement of these compounds with the antioxidant properties of the investigated honey samples from Bangladesh, as phenolic acids and flavonoids are well known for their antioxidant potential. Generally, these strong positive correlations suggest that the monofloral honey samples from Bangladesh have strong antioxidant potential.
# Conclusion
This is the first study to identify the phenolic compounds in honey samples from Bangladesh. Caffeic acid and benzoic acid were the most abundant phenolic compounds (70%) among all the phenolic acids, followed by kaempferol and gallic acid. In addition to these compounds, catechin, chlorogenic acid, myricetin, naringenin, and apigenin were also detected. Coniferic acid, naringenin, and hesperetin were not present in any of the investigated honey samples. The mean total phenolic acids, total flavonoid content, and proline content were 199.20, 46.73, and 556.40 mg/kg, respectively, while the mean FRAP values and DPPH radical-scavenging activity were 327.30 M Fe (II)/100g and 36.95%, respectively. Among the different types of honey, the honey samples collected from the Nigella sativa plant typically exhibited the highest phenolic content and antioxidant properties. A strong correlation was established between the color intensities of the honey samples and their antioxidant parameters: phenolic acids, flavonoids, and DPPH and FRAP values. However, further studies investigating on several different types of honey samples including Kalijira, Padma, and Teel honey are warranted. Overall, our study confirms that all the investigated honey samples from Bangladesh are good sources of phenolic acids and flavonoids, which confer their good antioxidant potential.
[fig] Figure 1: Sources of honey. [/fig]
[fig] Figure 2: Physical parameters such as (a) moisture content, (b) total sugar content, and (c) color intensity of the investigated honey samples. [/fig]
[fig] Figure 3, Figure 4: Color Total phenolic and flavonoid contents of the monofloral honey from Bangladesh. [/fig]
[fig] Figure 5, Figure 6: (a)), suggesting that they have different antioxidant potentials. Samples BH-3 and BH-10 again contained the highest FRAP values (786.22 M Fe (II)/100 g of honey) compared to the other samples. The mean FRAP value (327.30 M Fe (II)/100 g) of the investigated honey samples from Bangladesh is higher than that of acacia honey from Burkina Faso (79.5 M Fe (II)) and chestnut (388.6 M Fe (II)) and chicory honey (209.5 M Fe (II)) [7] from Slovenia, acacia honey [71.00 M Fe (II)] and lime honey [118.8 M Fe (II)] [3], suggesting the high quality of the honey samples from Bangladesh, as indicated by their high antioxidant potentials. (a) FRAP values and (b) proline content of monofloral honey from Bangladesh. DPPH radical-scavenging activities of the monofloral honey from Bangladesh. [/fig]
[fig] Figure 7: Typical chromatograms for (a) phenolic acid standards, (b) flavonoids standards, (c) phenolic acids and flavonoids in honey sample BH-3, and (d) phenolic acids and flavonoids in honey sample BH-6. GAL: gallic acid, CGN: chlorogenic acid, CAF: caffeic acid, CNF: coniferic acid, BNZ: benzoic acid, TCM: trans-cinnamic acid, CAT: catechin, NGI: naringenin, MYR: myricetin, NGE: naringenin, HSP: hesperitin, KMF: kaempferol, and APG: apigenin. [/fig]
[fig] Figure 8: Phenolic acids and flavonoid compounds detected in the different types of honey samples from Bangladesh using HPLC. [/fig]
[table] Table 1: Floral type and source of the investigated Bangladeshi monofloral honey. [/table]
|
Quasi-freestanding graphene on Ni(111) by Cs intercalation
A possible approach to achieve quasi-freestanding graphene on a substrate for technological purpose is the intercalation of alkali metal atoms. Cs intercalation between graphene and Ni(111) therefore is investigated using density functional theory, incorporating van der Waals corrections. It is known that direct contact between graphene and Ni(111) perturbs the Dirac states. We find that Cs intercalation restores the linear dispersion characteristic of Dirac fermions, which agrees with experiments, but the Dirac cone is shifted to lower energy, i.e., the graphene sheet is n-doped. Cs intercalation therefore decouples the graphene sheet from the substrate except for a charge transfer. On the other hand, the spin polarization of Ni(111) does not extend through the intercalated atoms to the graphene sheet, for which we find virtually spin-degeneracy.Graphene is gaining attention for its unique two-dimensional structure and special electronic properties, which are interesting for many applications 1 . A popular method to prepare graphene in high quality and large scale is chemical vapor deposition 2 , which has been successfully used for metallic substrates, such as Cu 3 , Au 4 , Co 5 , Ni 6 , Pd 7 , and Ir 8 . Ni(111) and Co(0001), in particular, are considered to be good substrates due to their small lattice mismatch of only 1.3% 9 and 1.9% 10 , respectively. However, some metallic substrates, such as Ni(111) 11 interact strongly with graphene and distort the Dirac cone 12 , whereas for Pt(111) 13 , for example, the interaction is weaker. The small distance (2.1 Å) realized between graphene and Ni(111) results in significant hybridization of the Ni 3d and C 2p orbitals, which explains the strong modification of the band structure 14,15 . It has been shown experimentally 16 and theoretically 17 that C magnetic moments are induced.Intercalation of noble metal atoms can restore the original band structure to give rise to quasi-freestanding graphene 18 . Ag intercalation into the graphene/Ni(111) system has been studied by density functional theory and it has been found that the Ag atoms decouple graphene from the substrate electronically 19 . The band gap decreases when the Ag concentration grows, as the interaction with the substrate is reduced. Ni then has no other role than to support the noble metal atoms. For Au intercalation the concentration also has a noticeable impact on the shape of the graphene π bands 20 . A low Au concentration of ~0.33 monolayers largely recovers the Dirac cone but leaves a band gap of 0.2 eV and a sizable shift of the Dirac point by 0.48 eV below the Fermi level. Increasing the Au coverage, on the other hand, almost removes the band gap and shifts the Dirac cone back to the Fermi level. Metal intercalation is not restricted to noble metals. In particular, alkali metals have been studied experimentally 11 and theoretically 21 . They also weaken the interaction between graphene and Ni(111) by enlarging the interlayer distance 22,23 .In the following, we consider the structural, electronic, and magnetic properties of graphene on Ni(111), using density functional theory, and analyze the changes when intercalating the system with different concentrations of Cs atoms. By the large atomic radius of Cs a very efficient decoupling of graphene from the substrate can be expected. Cs is an alkali metal with one electron in its outer shell and thus strongly favors a Cs +1 state, so that it does not participate in chemical bonding but acts purely as spacer.ResultsThe 1 × 1 unit cell of the graphene on Ni(111) consists of 4 Ni layers and 1 C layer on top. To determine the favorable lateral shift between the two materials, we study four structural configurations: (i) Hollow, where each atom of the topmost Ni layer is in the center of a C honeycomb. (ii) Face centered cubic (fcc), where three of the C honeycomb atoms are located on top of atoms of the topmost Ni layer and the center of the C honeycomb is located on top of an atom of the second Ni layer. (iii) Hexagonal close packed (hcp), which is similar to the fcc case but with the center of the C honeycomb located on top of an atom of the third Ni layer. (iv) Bridge, where the center of the C honeycomb is located between two atoms of the topmost Ni layer.Figure 1illustrates the four
[formula] Γ K M Γ K M -1 0 1 E-E F (eV) (b) (a) [/formula]
cases and summarizes key findings. In agreement with previous studies [bib_ref] Bridge Structure for the Graphene/Ni(111) System: A First Principles Study, Fuentes-Cabrera [/bib_ref] , the fcc configuration turns out to have the lowest formation energy where E Ni is the total energy per atom in bulk Ni and E C is the total energy per C atom in graphene. The fact that the bridge configuration is only 10 meV higher in energy than the fcc configuration implies that it is accessible as well. Even the hcp configuration might be accessible with low probability. Indeed, these configurations experimentally can coexist [bib_ref] Gold Intercalation of Boron-Doped Graphene on Ni(111): XPS and DFT Study, Zhao [/bib_ref] [bib_ref] Atomic Scale Identification of Coexisting Graphene Structures on Ni(111), Bianchini [/bib_ref]. The hollow configuration, on the other hand, is rather unlikely to be formed.
[formula] = − − E E E E ,(1)formation s ystem N i C Γ K M Γ K M -1 0 1 E-E F (eV) -10 -8 -6 -4 -2 0 2 E-E F (eV) [/formula]
The C-C bond length in the fcc configuration is 1.44 Å, whereas it is 1.42 Å in pristine graphene. This small tensile strain corresponds to the mentioned small lattice mismatch. The obtained perpendicular distance between graphene and the Ni(111) surface, d C/Ni , see , is in very close agreement with previous experimental [bib_ref] Atomic Structure of Monolayer Graphite Formed on Ni(111), Gamo [/bib_ref] and theoretical [bib_ref] Graphene on Ni(111): Strong Interaction and Weak Adsorption, Mittendorfer [/bib_ref] findings. For the fcc configuration, [fig_ref] Figure 2: Graphene on Ni [/fig_ref] shows the band structure for the spin up and down channels. The Dirac cone is strongly perturbed with a 0.3 eV splitting at the K point, reflecting a strong hybridization between the C 2p and Ni 3d states as a consequence of the small d C/Ni . We obtain finite though small magnetic We next consider a 2 × 2 supercell with graphene artificially placed in a distance of 6 Å from the Ni(111) substrate, see [fig_ref] Figure 3: Band structure of the [/fig_ref]. As expected, this distance yields a weak interaction between the Ni 3d and C 2p states, compare the almost vanishing Ni-C hybridization, leading to a restoration of the Dirac cone with a minor splitting. In fact, hybridization effects start vanishing in distances larger than 3 Å. [fig_ref] Figure 3: Band structure of the [/fig_ref] (a,b) shows the band structures of the spin up and down channels. We obtain a shift of the Dirac point by 0.2 eV above the Fermi level, reflecting p-doped graphene. The Dirac cone appears in the two spin channels at the same energy, implying that there is no C magnetic moment. The Ni bands, on the other hand, show a rigid shift due to spin polarization. Projected densities of states in [fig_ref] Figure 3: Band structure of the [/fig_ref] confirm that the magnetic moments on the C atoms are zero and that the total magnetic moment of 10.8 μ B comes from the Ni atoms, where atoms in the first layer show magnetic moments of 0.67 μ B and atoms in the other layers show 0.68 μ B .
[formula] Γ K M -1 0 1 E-E F (eV) Γ K M -10 -8 -6 -4 -2 0 2 E-E F (eV) [/formula]
We find that intercalation of Cs enlarges the distance between the graphene sheet and the Ni(111) substrate to 6.0 Å post relaxation (3.1 Å from graphene to Cs and 2.9 Å from Cs to Ni). The in-plane Cs-Cs distance is 5.0 Å. We note that the graphene sheet shows no significant structural modification, see [fig_ref] Figure 4: Band structure of the [/fig_ref] , in particular hardly any buckling. According to [fig_ref] Figure 4: Band structure of the [/fig_ref] , the electronic structure, on the other hand, is altered substantially by the Cs intercalation. In particular, a multitude of bands appears from − 1.5 to 0.5 eV. To understand the origin of these bands we use a weighted band analysis. [fig_ref] Figure 5: Atomic structures and weighted bands for the [/fig_ref] shows that the cone-like bands originate from the C 2p states, while the other bands are due to the Ni 3d states. The Dirac cone is restored with a negligible splitting. It appears about 1.1 eV below the Fermi level, representing prominently n-doped graphene. The C atoms carry no magnetic moment, see also [fig_ref] Figure 4: Band structure of the [/fig_ref] , implying that Cs does not transfer spin polarization from Ni to graphene. In the Ni(111) substrate the magnetic moments decrease towards Cs from a value of 0.68 μ B , which is close to the bulk Ni value of 0.67 μ B , to values of 0.49 μ B for the atom directly below Cs and 0.59 μ B for the other atoms in the same layer.
We next consider larger supercells, see [fig_ref] Figure 5: Atomic structures and weighted bands for the [/fig_ref] on top, to study the effect of the Cs concentration. The weighted band structures obtained for 2 × 2 (without and with intercalation), 3 × 3, and 4 × 4 supercells in [fig_ref] Figure 5: Atomic structures and weighted bands for the [/fig_ref] demonstrate a shift of the Dirac point upwards to the Fermi level when the Cs concentration is reduced. The higher the Cs concentration the stronger is the n-doping of graphene. The Dirac point is located 1.12 eV, 0.61 eV, and 0.59 eV below the Fermi level, respectively, for the 2 × 2, 3 × 3, and 4 × 4 supercell, which corresponds to a charge transfer of 0.045, 0.012, and 0.011 electrons in the case of pristine graphene.
As another ferromagnetic substrate we have also considered Co(0001) to check whether our results for Ni(111) are of general validity. The band structure of the 2 × 2 supercell, see [fig_ref] Figure 6: Band structure of the [/fig_ref] , is found to show close similarity to the case of the Ni(111) substrate, only the Dirac point is shifted slightly more (1.17 eV) below the Fermi level. The projected densities of states in [fig_ref] Figure 6: Band structure of the [/fig_ref] , on the other hand, indicate that Co is subject to a stronger exchange splitting as compared to Ni.
# Discussion
We have investigated the effect of Cs intercalation on the structural, electronic, and magnetic properties of graphene on Ni(111) and Co(0001). An fcc configuration is found to be energetically favorable in agreement with previous experimental and theoretical results, and therefore is chosen for studying the intercalation. Different supercell sizes have been considered to examine the effect of the Cs concentration. In agreement with experimental results, the Cs intercalation restores the Dirac cone, since it decouples graphene from the substrate, resulting in n-doped quasi-freestanding graphene. Both for the Ni(111) and Co(0001) substrates the graphene sheet exhibits no trace of spin polarization.
# Methods
Density functional theory is employed using the Vienna Ab-initio Simulation Package 30 and the generalized gradient approximation (Perdew-Burke-Ernzerhof parametrization 31 ) along with ultrasoft pseudopotentials. London forces are considered in all calculations 32 and the plane wave cutoff energy is set to 500 eV. Supercells are created using in the xy-plane the lattice parameter of Ni (2.49 Å) and adding a vacuum slab of about 15 Å thickness in the z-direction. For a 1 × 1 unit cell of graphene on Ni(111) a 32 × 32 × 1 k-mesh (Monkhorst-Pack scheme [bib_ref] Special Points for Brillouin-Zone Integrations, Monkhorst [/bib_ref] is used for the Brillouin zone integration (16 × 16 × 1 for the structure relaxation). We built a 2 × 2 supercell containing 24 atoms (8 C and 16 Ni) and introduce one Cs atom, giving a total number of 25 atoms. Also 3 × 3 and 4 × 4 supercells are constructed to examine the effect of the Cs concentration. These supercells are relaxed using 4 × 4 × 1 and 2 × 2 × 1 k-meshes, respectively, and 8 × 8 × 1 and 4 × 4 × 1 k-meshes are employed for the self-consistency calculation.
[fig] Figure 1: Top view of the structural configurations considered for graphene on Ni(111), where the C atoms (yellow) occupy (a) hollow, (b) fcc, (c) hcp, and (d) bridge sites on the Ni(111) surface. [/fig]
[fig] Figure 2: Graphene on Ni(111): Band structure of the (a) spin up and (b) spin down channels. [/fig]
[fig] Figure 3: Band structure of the (a) spin up and (b) spin down channels for d C/Ni = 6 Å. (c) Atomic structure with real and artificially exaggerated distance between the graphene sheet and Ni substrate. (d) Projected densities of states. Scientific RepoRts | 6:26753 | DOI: 10.1038/srep26753 [/fig]
[fig] Figure 4: Band structure of the (a) spin up and (b) spin down channels for the graphene/Cs/Ni(111) system (2 × 2 supercell). (c) Atomic structure with intercalated Cs (gray). (d) Projected densities of states. Scientific RepoRts | 6:26753 | DOI: 10.1038/srep26753 moments of − 0.02 and 0.03 μ B on the two C atoms due to Ni-C hybridization, in agreement with the results reported in ref. 29, and charge transfers of 0.03 and − 0.01 electrons from the Ni 3d to the C 2p states. [/fig]
[fig] Figure 5: Atomic structures and weighted bands for the (a,b) 2 × 2 supercell without/with intercalation, (c) 3 × 3 supercell, and (d) 4 × 4 supercell. The size of the dots represents the sum of all C p z contributions. Scientific RepoRts | 6:26753 | DOI: 10.1038/srep26753 [/fig]
[fig] Figure 6: Band structure of the (a) spin up and (b) spin down channels for the graphene/Cs/Co(0001) system (2 × 2 supercell). (c) Projected densities of states. Scientific RepoRts | 6:26753 | DOI: 10.1038/srep26753 [/fig]
|
Evaluating the Evidence for Lymphatic Filariasis Elimination
In the global drive for elimination of lymphatic filariasis (LF), 15 countries have achieved validation of elimination as a public health problem (EPHP). Recent empirical evidence has demonstrated that EPHP does not always lead to elimination of transmission (EOT). Here we show how the probability of elimination explicitly depends on key biological parameters, many of which have been poorly characterized, leading to a poor evidence base for the elimination threshold. As more countries progress towards EPHP it is essential that this process is wellinformed, as prematurely halting treatment and surveillance programs could pose a serious threat to global progress. We highlight that refinement of the weak empirical evidence base is vital to understand drivers of elimination and inform long-term policy.Global Situation and ProgressThere are currently 886 million people across 52 countries worldwide at risk of LF i . Infection is caused by a mosquito-transmitted filarial worm and, if left untreated, can lead to permanent and debilitating disability. The Global Program to Eliminate Lymphatic Filariasis (GPELF) set a target of elimination as a public health problem (EPHP) (see Glossary) in 1997, leading to over 7.1 billion treatments delivered as part of mass drug administrations (MDAs) since 2000 i . In 2011, the WHO published guidelines for halting treatment and verifying EPHP through the use of transmission assessment surveys (TAS) to measure a target threshold ii,iii . By October 2018, 14 countries had reached this target, and 554 million people worldwide no longer require mass treatments iv .As indicated by the name of the TAS, it was hoped that reaching these targets would lead to elimination of transmission (EOT) in most areas. However, in Sri Lanka the TAS has been demonstrated as not sensitive enough to detect low-level persistence [1,2], and pockets of transmission are still being found despite EPHP validation. The community is now revisiting the TAS methods, including the original target of 1% microfilaria (mf) prevalence [3], particularly in the context of the new triple-drug regimen which is hoped to accelerate progress, but will require different post-treatment surveillance[4].It is possible that achieving EPHP, according to the current definition, will lead to EOT in some settings[5,6], but the high levels of variability between localities, and uncertainty in our knowledge of transmission, make it hard to predict where this will occur. This is exacerbated further by seasonal variation in environmental conditions, which has been shown to impact a number of helminth infections[7,8]. Residual infection remaining after MDA cessation can lead to resurgence and reintroduction [9,10], with long-term persistence dependent on a range of factors[11].Sexual Reproduction in the Host, and EliminationThe sexual reproduction of filarial worms requires both male and female parasites to be present in an individual host for microfilariae production, so at a sufficiently low prevalence we would expect most infections to be nontransmissible due to low parasite load (i.e., a low probability of male and female adults in the same host). This is expected to result in fewer onward infections, and hence increasingly lower prevalence and intensity, until infection dies out. The threshold below which we expect this phenomenon to occur is called the breakpoint [12,13]. As the focus of some neglected tropical disease (NTD) programs has shifted from control towards elimination, there have been a number of studies aiming to quantify these thresholds for a variety of helminth infections within the NTD umbrella[14][15][16][17].
In the global drive for elimination of lymphatic filariasis (LF), 15 countries have achieved validation of elimination as a public health problem (EPHP). Recent empirical evidence has demonstrated that EPHP does not always lead to elimination of transmission (EOT). Here we show how the probability of elimination explicitly depends on key biological parameters, many of which have been poorly characterized, leading to a poor evidence base for the elimination threshold. As more countries progress towards EPHP it is essential that this process is wellinformed, as prematurely halting treatment and surveillance programs could pose a serious threat to global progress. We highlight that refinement of the weak empirical evidence base is vital to understand drivers of elimination and inform long-term policy.
## Global situation and progress
There are currently 886 million people across 52 countries worldwide at risk of LF i . Infection is caused by a mosquito-transmitted filarial worm and, if left untreated, can lead to permanent and debilitating disability. The Global Program to Eliminate Lymphatic Filariasis (GPELF) set a target of elimination as a public health problem (EPHP) (see Glossary) in 1997, leading to over 7.1 billion treatments delivered as part of mass drug administrations (MDAs) since 2000 i . In 2011, the WHO published guidelines for halting treatment and verifying EPHP through the use of transmission assessment surveys (TAS) to measure a target threshold ii,iii . By October 2018, 14 countries had reached this target, and 554 million people worldwide no longer require mass treatments iv .
As indicated by the name of the TAS, it was hoped that reaching these targets would lead to elimination of transmission (EOT) in most areas. However, in Sri Lanka the TAS has been demonstrated as not sensitive enough to detect low-level persistence [bib_ref] A comprehensive assessment of lymphatic filariasis in Sri Lanka six years after..., Rao [/bib_ref] [bib_ref] Elimination of lymphatic filariasis in The Gambia, Rebollo [/bib_ref] , and pockets of transmission are still being found despite EPHP validation. The community is now revisiting the TAS methods, including the original target of 1% microfilaria (mf) prevalence, particularly in the context of the new triple-drug regimen which is hoped to accelerate progress, but will require different post-treatment surveillance [bib_ref] Effectiveness of a triple-drug regimen for global elimination of lymphatic filariasis: a..., Irvine [/bib_ref].
It is possible that achieving EPHP, according to the current definition, will lead to EOT in some settings [bib_ref] The history of the elimination of lymphatic filariasis in China, De-Jian [/bib_ref] [bib_ref] Use of antibody tools to provide serologic evidence of elimination of lymphatic..., Won [/bib_ref] , but the high levels of variability between localities, and uncertainty in our knowledge of transmission, make it hard to predict where this will occur. This is exacerbated further by seasonal variation in environmental conditions, which has been shown to impact a number of helminth infections [bib_ref] Seasonally timed treatment programs for Ascaris lumbricoides to increase impact -an investigation..., Davis [/bib_ref] [bib_ref] Modelling the impact of vector control interventions on Anopheles gambiae population dynamics, White [/bib_ref]. Residual infection remaining after MDA cessation can lead to resurgence and reintroduction [bib_ref] GEOFIL: A spatially-explicit agentbased modelling framework for predicting the longterm transmission dynamics..., Xu [/bib_ref] [bib_ref] Bayesian calibration of simulation models for supporting management of the elimination of..., Singh [/bib_ref] , with long-term persistence dependent on a range of factors [bib_ref] Elimination within reach: A cross-sectional study highlighting the factors that contribute to..., Minetti [/bib_ref].
## Sexual reproduction in the host, and elimination
The sexual reproduction of filarial worms requires both male and female parasites to be present in an individual host for microfilariae production, so at a sufficiently low prevalence we would expect most infections to be nontransmissible due to low parasite load (i.e., a low probability of male and female adults in the same host). This is expected to result in fewer onward infections, and hence increasingly lower prevalence and intensity, until infection dies out. The threshold below which we expect this phenomenon to occur is called the breakpoint [bib_ref] Assessing the interruption of the transmission of human helminths with mass drug..., Anderson [/bib_ref]. As the focus of some neglected tropical disease (NTD) programs has shifted from control towards elimination, there have been a number of studies aiming to quantify these thresholds for a variety of helminth infections within the NTD umbrella [bib_ref] Identifying optimal threshold statistics for elimination of hookworm using a stochastic simulation..., Truscott [/bib_ref] [bib_ref] Six challenges in the eradication of infectious diseases, Klepac [/bib_ref] [bib_ref] Quantitative analyses and modelling to support achievement of the 2020 goals for..., Hollingsworth [/bib_ref].
The current target of elimination as a public health problem (EPHP) for lymphatic filariasis was originally devised with the intention of interrupting transmission. However, some countries that have achieved EPHP are still finding new cases.
Analysis of the evidence for key biological determinants suggests that a target threshold of <1% microfilaria (mf) prevalence is not likely to be sufficient for transmission interruption in communities with a mid-to-high annual biting rate.
The experimental evidence underlying estimates is insufficient or inconsistent, particularly transmission rates from vector to human, leading to high uncertainty in confidence of elimination success.
Local biting rate is expected to be highly variable between settings and could have a large impact on elimination feasibility for a given target prevalence.
Further experimental studies are needed to refine our understanding of LF elimination thresholds. Basic reproductive number (R 0 ): the average number of new infectious cases generated by one infectious case in an entirely susceptible population. Blood feeding rate (BFR): the rate at which mosquitoes take a blood meal.
## Trends trends in in parasitology parasitology
Branching process: a stochastic process which consists of collections of random variables, which are indexed by the natural numbers (1,2,3,.). Breakpoint: a prevalence level below which sustained transmission is not viable and elimination (zero cases) becomes an absorbing state. Effective reproductive number (R e ): the average number of new infectious cases generated by one infectious case in a population made up of both susceptible and infectious hosts. Elimination as a public health problem (EPHP): as measured by TAS, a metric used by the WHO to validate programme success. Intended to naturally lead to EOT. Extrinsic incubation period (EIP): the time it takes for ingested mf to develop to infectious L3 larvae in the mosquito. Implementation unit: the designated level of the administrative unit in a country, for which the decision to administer antifilarial drugs to the entire population is taken if it is identified as having indigenous transmission or endemicity. This theory has certain consequences for control [fig_ref] Figure 1: Lymphatic Filariasis Extinction Theory [/fig_ref]. If transmission is sufficiently low, then the infection is expected to die out. If there is a higher transmission rate, outcomes depend on the mean worm load in the population; if, usually through control strategies, the worm load is below the green broken line (the breakpoint) then elimination is assured. Previous modelling studies that have assessed breakpoint thresholds have found values of much less than 1% mf prevalence [bib_ref] Bayesian calibration of simulation models for supporting management of the elimination of..., Singh [/bib_ref] [bib_ref] Continental-scale, datadriven predictive assessment of eliminating the vector-borne disease, lymphatic filariasis, in..., Michael [/bib_ref] [bib_ref] Substantiating freedom from parasitic infection by combining transmission model predictions with disease..., Michael [/bib_ref] [bib_ref] Geographic and ecologic heterogeneity in elimination thresholds for the major vector-borne helminthic..., Gambhir [/bib_ref]. It has been previously demonstrated that factors such as parasite aggregation and vector competence will further affect these thresholds [bib_ref] Modelling strategies to break transmission of lymphatic filariasis -aggregation, adherence and vector..., Irvine [/bib_ref] , and the majority of studies have focused on specific geographical areas, resulting in a wide range of suggested breakpoints across the literature.
Measuring breakpoints that are substantially lower than 1% mf prevalence would require infeasible sample sizes and survey costs. In this review we do not argue for a specific breakpoint, instead focusing on asserting that the experimental evidence is too uncertain to conclusively support a 1% threshold and emphasizing the importance of spatial heterogeneity.
Whilst breakpoint theory is extremely useful, it is also possible for stochastic, or chance, extinction to occur before this breakpoint is reached, particularly when infection levels are low [fig_ref] Figure 1: Lymphatic Filariasis Extinction Theory [/fig_ref]. The probability of elimination, given a particular prevalence (e.g., 1%), can be calculated by considering the probability that a chain of transmission will die out (in mathematics we call this chain a branching process [bib_ref] On the probability of the extinction of families, Watson [/bib_ref]. These types of branching process methods have been used for soil-transmitted helminths [bib_ref] Stochastic and spatial dynamics of nematode parasites in farmed ruminants, Cornell [/bib_ref] [bib_ref] Ultimate extinction of the promiscuous bisexual Galton-Watson, Cornell [/bib_ref] , but have been adapted here to account for vector-borne transmission with an aggregated bite risk [bib_ref] The dynamics of Ascaris lumbricoides infections, Fowler [/bib_ref] [bib_ref] Mosquito bite heterogeneity influences lymphatic filariasis prevalence, intensity and opportunities for control, Irvine [/bib_ref].
Current guidelines mean that EPHP is validated after passing TAS, but we have little experience in what this means for long-term transmission. Assuming for simplicity that TAS is able to measure a true mf prevalence of less than 1%, this theory of stochastic extinction can be used to estimate how the future probability of EOT (zero cases) depends on a range of setting-and disease-specific variables. This process uses the distribution of the number of infectious secondary cases caused by one infectious individual, the mean of which is the effective reproductive number (R e ).
As a toy example, for a population of 1000 and 1% mf prevalence, we consider a distribution of individual worm burdens [fig_ref] Figure 2: Simulating Branching Process Extinction [/fig_ref]. Infections with only one worm are nontransmissible. From one infectious person you then get the number of new cases, Z, caused during their infectious period [fig_ref] Figure 2: Simulating Branching Process Extinction [/fig_ref]. Since transmission represents a chance event, Z is best represented by a distribution, and acts as a proxy for R e . This distribution determines the probability of the transmission chain dying out, that is, no further cases, at some point in the future; for more detail see Box S1 in the supplemental information online. We use this to give a univariate demonstration of the present parameter uncertainty and how this might impact two epidemiological measures: the probability of elimination and the effective reproductive number.
## Empirical evidence for life-cycle variables
We now review evidence for key parameters in the life cycle which drive transmission [fig_ref] Figure 3: Lymphatic Filariasis Life-cycle [/fig_ref]. As previously mentioned, a number of these variables, such as the annual biting rate (ABR), are likely to introduce large differences due to the high spatial variability. Others, such as the probability an infectious mosquito bite results in a viable human infection, have the potential to be more consistent across settings, but currently lack in experimental evidence.
A detailed literature review turns up widely varying estimates of ABR, partially due to geographical variation. These values, from countries with a history of LF endemicity, range from three [bib_ref] A simplified model for predicting malaria entomologic inoculation rates based on entomologic..., Killeen [/bib_ref] to 611 that has recently become the gold standard for treatment of LF. Univariate: literally 'of one variable'. Univariate analysis explores variables one-by-one, keeping all others fixed. Vector-host ratio: the number of vectors per human in a geographical region.
will increase to the higher equilibrium, whereas disease levels below will decrease to zero. (B) Visual depiction of a branching process starting with one infectious individual. The number of secondary infections caused by each currently infectious individual are sampled from the secondary case distribution. This is used to simulate the onward chains of infection; extinction occurs when all chains die out (i.e., have no secondary cases). Stochastic variation can cause this to occur even above the theoretical breakpoint threshold. [bib_ref] Heterogeneous dynamics, robustness/fragility trade-offs, and the eradication of the macroparasitic disease, lymphatic..., Michael [/bib_ref] bites per person per day. A number of these are based on human landing catches [bib_ref] A simplified model for predicting malaria entomologic inoculation rates based on entomologic..., Killeen [/bib_ref] [bib_ref] Heterogeneous dynamics, robustness/fragility trade-offs, and the eradication of the macroparasitic disease, lymphatic..., Michael [/bib_ref] [bib_ref] Biting behaviour of African malaria vectors: 1. where do the main vector..., Braack [/bib_ref] , with the majority relying on studies from the 1960s and 1970s [bib_ref] Heterogeneous dynamics, robustness/fragility trade-offs, and the eradication of the macroparasitic disease, lymphatic..., Michael [/bib_ref] , whilst some are derived from models. Despite a wealth of historic studies, supported by the malaria literature, human landing catches are often considered unethical and give highly variable results. Relying on historic estimates can also disregard changes in socioeconomic conditions resulting in decreased vector-human contact.
Current estimates in the literature of the basic reproductive number, R 0 , range from zero to 2.5 [bib_ref] How effective is integrated vector management against malaria and lymphatic filariasis where..., Stone [/bib_ref] , depending on the vector-host ratio (an alternative metric to ABR). Although setting-specific values of R 0 for other diseases can often be calculated from infection data, the global landscape of public health history for LF means that we have very little contemporary baseline (precontrol) data with which to do this. As an alternative, we can consider the previously mentioned estimation of R e . Another important, but largely uncertain, factor is the degree of parasite aggregation, measured inversely by the negative binomial k. For LF, adult worm aggregation is considered to be driven by heterogeneous transmission, caused by host variation in bite risk. Initial estimates for k were based on mf data (k = 0.08, 0.3 [bib_ref] Modelling strategies to break transmission of lymphatic filariasis -aggregation, adherence and vector..., Irvine [/bib_ref] [bib_ref] Mosquito bite heterogeneity influences lymphatic filariasis prevalence, intensity and opportunities for control, Irvine [/bib_ref]. However, a recent study in Papua New Guinea used bite and mf data to demonstrate that the k for bite risk is an order of magnitude larger than that for mf aggregation, giving a refined estimate of 0.73 (standard deviation 0.035), with site-specific estimates ranging from 0.3 to 1.3 [bib_ref] Mosquito bite heterogeneity influences lymphatic filariasis prevalence, intensity and opportunities for control, Irvine [/bib_ref]. We will now separate transmission into two parts: humans to mosquitoes, and mosquitoes to humans. When considering the former, the key variables are duration of infection, which depends on fecund worm lifespan, and the probability that a vector biting an infected host will become infectious.
## (b) (a)
## Trends trends in in parasitology parasitology
Often worm lifespan is stated as being 6-8 i or 5-10 years [bib_ref] EPIFIL: The development of an age-structured model for describing the transmission dynamics..., Norman [/bib_ref] [bib_ref] Diagnosis of lymphatic filariasis infection: How many people have adult worms? Am, Stolk [/bib_ref] , but reference trails rarely reveal empirical evidence. There are studies that corroborate similar ranges, such as 2.1-5.4 [bib_ref] Estimation of the fecund life span of Wuchereria bancrofti in an endemic..., Vanamail [/bib_ref] or 9.1-11.8 [bib_ref] The dynamics of Wuchereria bancrofti infection: a model-based analysis of longitudinal data..., Subramanian [/bib_ref] years, but there are also estimates in the literature of up to 40 years [bib_ref] Longevity of Wuchereria bancrofti var Pacifica and mosquito infection acquired from a..., Carme [/bib_ref]. Life-cycle schematic demonstrating key biological variables that could affect prediction of elimination success. Duration of infection is determined by human and fecund worm lifespans. Infection from host to vector depends on the annual biting rate (ABR) and the probability that a bite on an infectious host infects a vector. The number of vectors that survive to infectivity depends on the extrinsic incubation period (EIP) and vector lifespan. Transmission from vector to host is then determined by the blood feeding rate and the probability that an infectious bite results in a viable adult infection, as well as the requirement for two or more worms for infectivity.
## Probability bite on abr
## Trends trends in in parasitology parasitology
Infectivity to mosquitoes depends on mf intensity, leading to wide ranges of 15-60% of vectors becoming infected from a single mf-positive bite [bib_ref] The relationship between microfilarial load in the human host and uptake and..., Subramanian [/bib_ref] [bib_ref] Complex ecological dynamics and eradicability of the vector borne macroparasitic disease, lymphatic..., Gambhir [/bib_ref].
Infection from vector to human is governed by the number of infectious bloodmeals one mosquito will take -calculated from vector survival and competence, extrinsic incubation period (EIP) and blood feeding rate (BFR) -and the probability one infectious bite will result in a viable infection. There are reasonable estimates for vector survival and BFR from the malaria literature [bib_ref] A simplified model for predicting malaria entomologic inoculation rates based on entomologic..., Killeen [/bib_ref] [bib_ref] Effects of available sugar on the reproductive fitness and vectorial capacity of..., Gary [/bib_ref] and for LF incubation [bib_ref] Mosquito infection responses to developing filarial worms, Erickson [/bib_ref] , although these do not typically account for the impact of infection on survival [bib_ref] The relationship between microfilarial load in the human host and uptake and..., Subramanian [/bib_ref].
One key parameter of infection, the probability an infectious bite results in a mature human infection, is largely unknown. Estimates range from 10 -5 to 10 -3 [bib_ref] On the inefficiency of transmission of Wuchereria bancrofti from mosquito to human..., Hairston [/bib_ref] [bib_ref] Non-endemic cases of lymphatic filariasis, Jones [/bib_ref] and are usually broken down into three steps: the L3 leaving the vector, entering the host, and developing to fecundity. The first step is relatively straightforward to measure [bib_ref] Infection and reinfection of Culex pipiens fatigans with Wuchereria bancrofti and the..., De Meillon [/bib_ref] , although it poses ethical issues, and the second can be estimated using mouse models [bib_ref] Experimental transmission of filarial larvae in relation to feeding behaviour of the..., Ho [/bib_ref] [bib_ref] The fate of Brugia pahangi larvae immediately after feeding by infective vector..., Ewert [/bib_ref]. The third is harder; best estimates are calculated by using Brugia malayi studies to derive a daily death rate and then applying this across the longer Wucheria bancrofti developmental period [bib_ref] EPIFIL: The development of an age-structured model for describing the transmission dynamics..., Norman [/bib_ref].
## Quantifying the probability of elimination
If we include these parameters in the simple framework described above, we can see how the uncertainty affects our estimates of key epidemiological measures [fig_ref] Figure 4: Predicting Elimination Probabilities [/fig_ref]. The mid-points of elimination probability (0.73) and R e (1.1) are not intended to be true estimates, rather they represent a midground of the parameter ranges found in the literature and a basis for comparison.
The variable which generates the most univariate uncertainty is the probability that an infectious mosquito bite will infect a human, b, due to the wide range of possible values. Variation in elimination probability due to ABR, which is correlated with the basic reproductive number (R 0 ), is also very high. This is due to both measurement inaccuracy and spatiotemporal variability. Parameters that are known to be key drivers in the probability of elimination, worm fecund lifespan and the degree of adult worm aggregation [bib_ref] Modelling strategies to break transmission of lymphatic filariasis -aggregation, adherence and vector..., Irvine [/bib_ref] [bib_ref] Soil-transmitted helminths: mathematical models of transmission, the impact of mass drug administration..., Truscott [/bib_ref] [bib_ref] The coverage and frequency of mass drug administration required to eliminate persistent..., Anderson [/bib_ref] , potentially induce lower uncertainty here due to considering narrower plausible intervals.
In addition to the probability of elimination, we also consider the effective reproductive number, R e . It is important to note that, for helminth infections, metrics often refer to the number of adult filarial worms arising from one adult filarial worm, rather than considering human cases. However, the theory is similar enough to allow heuristic comparison. Our mid-estimate for R e is chosen to be close to 1, representative of the low-level transmission observed in some post-MDA settings, but varying the probability that an infectious mosquito bite will lead to a patent infection (b) can lead to an order of magnitude difference. In fact, it is possible to push the estimate of R e across the critical threshold (R e = 1) between extinction and endemicity by adjusting any variable within the ranges found in the literature. This reinforces the importance of using reliable variable estimates when making predictions, particularly in elimination settings where infection data are sparse.
## Recommendations
Due to the demonstrated uncertainty that knowledge gaps, particularly in the establishment of a patent infection, can cause in estimating elimination thresholds it would be prudent to refine the evidence for these variables. Here we discuss a few options for future studies and analyses that we believe could strengthen the knowledge base.
The probability that an infectious bite leads to an infectious host cannot be measured experimentally in humans; however, we can improve current estimates with anecdotal and observational studies. Longitudinal studies can provide evidence of the time to antigen positivity and the time to microfilaria positivity in children, or in adults that have moved from nonendemic to endemic regions. One existing study, looking at acquisition in travelers, surmises that the majority of cases are in individuals who spent in excess of 6 months in an endemic region [bib_ref] Filariasis in travelers presenting to the GeoSentinel Surveillance Network, Lipner [/bib_ref] , whereas another cites a number of travelers contracting infection with only 1 month of exposure [bib_ref] Non-endemic cases of lymphatic filariasis, Jones [/bib_ref]. Entomological studies routinely estimate ABR through human landing catch data, and individual exposure can be quantified based on net usage and vector biting habits [bib_ref] Insecticidal bed nets and filariasis transmission in Papua New Guinea, Reimer [/bib_ref] [bib_ref] Mosquito behavior change after distribution of bednets results in decreased protection against..., Thomsen [/bib_ref].
The range of ABRs discussed are very broad estimates, covering a wide range of settings, but this can be a difficult variable to measure consistently. It may be possible to obtain greater certainty in R e without accurate ABR measures for each location. For example, estimates of low, medium, or high vector densities would still improve our predictions, and these categories of exposure, which act as a proxy for R 0 classification, could be informed by a combination of trap densities and vector-control coverage. Spatial heterogeneity can also occur within implementation units, posing problems for any categorization process, so it is important that treatment targets are determined by the maximum transmission measure for a region.
## Concluding remarks
We have used basic analyses to highlight that the existing experimental evidence does not afford a high degree of certainty at the current 1% mf prevalence elimination threshold. This is mainly because of uncertainties in variables which could be either experimentally or analytically refined, but also due to spatiotemporal variation in vector densities and biting rates [bib_ref] Heterogeneous dynamics, robustness/fragility trade-offs, and the eradication of the macroparasitic disease, lymphatic..., Michael [/bib_ref]. That varying the value of one input variable within sensible ranges found in the literature can make such an impact on predictions, demonstrates the difficulties posed by targeting EOT when we know that local heterogeneities and variability are difficult to measure. Observations of ongoing transmission in parts of validated countries offer empirical support to our concerns with the EPHP target, prompting some important outstanding policy questions (see Outstanding Questions).
In order to support efforts to eliminate LF we would recommend a multipronged approach: improving the experimental evidence base of measurable quantities; detailed analysis of existing infection data to improve our understanding of the infection risk associated with an infectious bite; and development of a discrete system to classify vector density, as a proxy for transmission intensity, to allow comparison of different regions. The optimization of elimination programme strategies and surveillance will require continual revisiting of predictions as we gather more epidemiological data through existing surveys and monitoring infrastructures, as well as expanded epidemiological and surveillance studies at low prevalence.
As more countries cease interventions and move to postvalidation surveillance it is increasingly obvious that transmission breakpoints are unlikely to be one-size-fits-all, hence more flexible thresholds are necessary. It is vital that we ensure that this process is well-informed, as prematurely halting control or surveillance programs could pose a serious threat to global targets, but also because we believe that it may be possible to exploit this geographical variation to maximize the probability of elimination. Illustration of the potential impact of high uncertainty in variables by considering their univariate impact on the probability of elimination (A) and the effective reproductive number (B) for the key biological variables of the lymphatic filariasis (LF) transmission cycle, assuming a microfilaria (mf) prevalence of 1% and a human population size of 1000. References for ranges of variables considered can be found in in the supplemental information online. Note that this univariate analysis should be interpreted carefully as variables are likely to be correlated in ways which we cannot yet account for. For example, the mid-estimates here have been chosen to represent a mid-ground of ranges found in the literature and are not necessarily representative of the true values or ranges that may exist across real-world settings. Abbreviations: BFR, blood feeding rate; EIP, extrinsic incubation period.
## Outstanding questions
How can we translate our understanding of elimination dynamics to clear and feasible guidelines for public health programs?
Is there a universal threshold, or do we need to tailor predictions for different communities and settings?
What are the key determinants that vary between settings, and how can we measure them?
How can we reliably measure annual biting rate for different settings?
How can we refine our estimates of transmission probability from vectors to humans?
How can we determine where 1% mf prevalence is a threshold below which elimination is likely?
If lower target thresholds are required for elimination of transmission, then are we realistically able to measure these using current tools?
Whilst we can measure that prevalence is below certain thresholds, is this sufficient evidence of elimination of transmission?
In settings where we are still seeing new cases after EPHP verification, what is the probability of largescale resurgence?
What is a suitable survey design in a context of limited resources?
How will the new diagnostic affect elimination measurement?
How can we harness xeno-monitoring techniques to improve post-EPHP surveillance?
## Supplemental information
Supplemental information associated with this article can be found online at https://doi.org/10.1016/ j.pt.2019.08.003.
Resources i www.who.int/lymphatic_filariasis/resources/who_wer9344/en/ ii www.who.int/lymphatic_filariasis/resources/9789241511957/en/ iii www.who.int/lymphatic_filariasis/resources/9789241550161/en/ iv www.who.int/news-room/fact-sheets/detail/lymphatic-filariasis
[fig] Figure 1: Lymphatic Filariasis Extinction Theory. Schematics comparing the theory behind breakpoint extinction (A) and stochastic extinction (B) for lymphatic filariasis. (A) For sufficiently low transmission intensities (i.e., low biting rates), disease levels will drop away to zero. Beyond the critical transmission level (black broken line) there are three equilibria: high disease (stable, red), low disease (unstable 'breakpoint', green), disease-free (stable, black). Disease levels above the breakpoint Glossary Annual biting rate (ABR): the average number of mosquito bites per person per year. [/fig]
[fig] Figure 2: Simulating Branching Process Extinction. A schematic describing the simulation process for calculating the number of secondary cases produced by one infectious individual in a population with 1% microfilaria (mf) prevalence. (A) Allocate distribution of adult worms and bite risks across the population. Individuals with 1 worm are infected but are not infectious, individuals with two or more worms are considered potentially infectious. (B) Generational calculation of number of new infectious cases caused by one infectious individual. One infectious individual infects X vectors. The vectors that survive the incubation take infectious blood meals, resulting in Y new adult worms. These worms are distributed across the population according to bite risk aggregation, resulting in Z new infectious (R2 worms) individuals. [/fig]
[fig] Figure 3: Lymphatic Filariasis Life-cycle. [/fig]
[fig] Figure 4: Predicting Elimination Probabilities. [/fig]
|
Collagens and Cancer associated fibroblasts in the reactive stroma and its relation to Cancer biology
The extracellular matrix (ECM) plays an important role in cancer progression. It can be divided into the basement membrane (BM) that supports epithelial/endothelial cell behavior and the interstitial matrix (IM) that supports the underlying stromal compartment. The major components of the ECM are the collagens. While breaching of the BM and turnover of e.g. type IV collagen, is a well described part of tumorigenesis, less is known regarding the impact on tumorigenesis from the collagens residing in the stroma. Here we give an introduction and overview to the link between tumorigenesis and stromal collagens, with focus on the fibrillar collagens type I, II, III, V, XI, XXIV and XXVII as well as type VI collagen. Moreover, we discuss the impact of the cells responsible for this altered stromal collagen remodeling, the cancer associated fibroblasts (CAFs), and how these cells are key players in orchestrating the tumor microenvironment composition and tissue microarchitecture, hence also driving tumorigenesis and affecting response to treatment. Lastly, we discuss how specific collagen-derived biomarkers reflecting the turnover of stromal collagens and CAF activity may be used as tools to non-invasively interrogate stromal reactivity in the tumor microenvironment and predict response to treatment.
# Introduction
The ECM is an extensive part of the microenvironment in all tissues. It consists of a non-cellular meshwork of proteins, glycoproteins, proteoglycans and polysaccharides. When structured in an orderly manner, the ECM provides a physical scaffold for its surrounding cells, bind growth factors and regulate cell behavior.
The ECM can be divided into two matrices: the basement membrane (BM) and the interstitial matrix (IM). Under healthy conditions, the BM is a well-structured membrane underlining epithelial and endothelial cells and separating them from the IM. When fully assembled the BM provides structural support to underlining cells and regulate cell behavior. The IM makes up the main stroma and plays a major role in cell migration, cell adhesion, angiogenesis, tissue development and repair [bib_ref] Extracellular matrix structure, Theocharis [/bib_ref].
The major proteins in the ECM are collagens, which constitutes up to 30% of the total protein mass in the human body [bib_ref] The extracellular matrix at a glance, Frantz [/bib_ref]. The collagens are organized in a relaxed meshwork surrounded by proteins such as elastin and glycoproteins causing a resilience to extensive tensile strength [bib_ref] The extracellular matrix at a glance, Frantz [/bib_ref]. Of today, 28 different collagens have been identified creating a unique ECM composition in different tissues. The 28 collagens can be divided into several distinct subgroups, where the so-called fibrillar-forming collagens and the network-forming collagens have been most extensively characterized. The major components of the BM are the network-forming collagens such as type IV and type VIII collagen whereas the IM is dominated by the fibrillar-forming collagens type I, II, III, V, XI, XXIV, XXVII and the beaded filament type VI collagen synthesized by the fibroblasts receding in the stroma [bib_ref] Type I collagen synthesis in cultured human fibroblasts: regulation by cell spreading,..., Ivarsson [/bib_ref] [bib_ref] Simultaneous synthesis of types I and III collagen by fibroblasts in culture, Gay [/bib_ref] [bib_ref] Synthesis of type V collagen by fibroblasts derived from Normal, inflamed and..., Sampath Narayanan [/bib_ref] [bib_ref] Muscle interstitial fibroblasts are the main source of collagen VI synthesis in..., Zou [/bib_ref] [bib_ref] Characterization of type II and type XI collagen synthesis by an immortalized..., Oxford [/bib_ref]. These collagens are not just collagens but individual structures creating a complex network that interact with each other and the surroundings [fig_ref] Figure 1: Collagens within the basement membrane and interstitial matrix [/fig_ref].
In the healthy tissue there is an ongoing ECM remodeling to maintain tissue integrity and function, e.g. new collagens synthesized that replaces older proteins that are degraded. The collagen production and assembly in healthy tissue is highly regulated by a perfect counterbalance of metalloproteinases (MMPs) and inhibitors of MMPs as well as a controlled activity of other enzymes such as lysyl oxidases (LOX) [bib_ref] Remodelling the extracellular matrix in development and disease, Bonnans [/bib_ref] [bib_ref] Proteolytic networks in Cancer, Mason [/bib_ref].
During cancer, the ECM-dynamics are skewed. It is well established that cancer cells secrete high amounts of MMPs, which in turn remodel and degrade the BM. The remodeling of the BM leads to a complex chaos of pro-and antitumor signals from degradation products. The role of type IV collagen turnover, within the BM, has been extensively studied in relation to tumor biology. Several studies have shown that proteolytic cleavage of collagen IV can expose so-called cryptic domains, which are normally hidden when collagen IV is fully assembled [bib_ref] Basement membranes: structure, assembly and role in tumour angiogenesis, Kalluri [/bib_ref] [bib_ref] The role of type IV collagen and basement membranes in cancer progression..., Tanjore [/bib_ref] [bib_ref] Proteolytic exposure of a cryptic site within collagen type IV is required..., Xu [/bib_ref] [bib_ref] Generation of monoclonal antibodies to cryptic collagen sites by using subtractive immunization, Xu [/bib_ref]. Similar things have been seen with other BM collagens e.g. type XVIII collagen [bib_ref] Endostatin: an endogenous inhibitor of angiogenesis and tumor growth, O'reilly [/bib_ref]. Depending on the context, these cryptic sites have both pro-and anti-tumor effects; still the turnover and degradation of BM collagens are intrinsically associated with the invasive phenotype of malignant cells [bib_ref] Basement membranes: structure, assembly and role in tumour angiogenesis, Kalluri [/bib_ref]. V, XI, XXIV and XXVII collagens are all fibrillary collagens embedded in the IM, whereas type VI collagen is found in the interface between the BM and the IM. Emerging evidence indicate a high impact of fibroblast-derived collagens and so-called cancer associated fibroblasts (CAFs) in tumorigenesis [bib_ref] The biology and function of fibroblasts in cancer, Kalluri [/bib_ref] [bib_ref] Fibroblasts in cancer, Kalluri [/bib_ref]. During tumor progression, CAFs are the major players in the dysregulated collagen turnover leading to tumor fibrosis (desmoplasia) characterized by excessive collagen depositions in the surroundings of the tumor [bib_ref] Cancer-associated fibroblasts induce a collagen cross-link switch in tumor stroma, Pankova [/bib_ref]. The collagens are often crosslinked and linearized leading to increased stiffening of the tissue . This elicits behavioral effects on surrounding tumor cells, and regulate cell proliferation, differentiation, gene expression, migration, invasion, metastasis and survival and hereby the collagens are directly affecting the hallmarks of cancer [bib_ref] The extracellular matrix modulates the hallmarks of cancer, Pickup [/bib_ref]. In support, tumor tissue, containing a large amount of these fibroblast derived stromal collagens is directly correlated with poorer outcome for the patient [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Extracellular matrix specific protein fingerprints measured in serum can seperate pancreatic cancer..., Willumsen [/bib_ref] [bib_ref] Serum biomarkers reflecting specific tumor tissue remodeling processes are valuable diagnostic tools..., Willumsen [/bib_ref] [bib_ref] Excessive collagen turnover products are released during colorectal cancer progression and elevated..., Kehlet [/bib_ref] [bib_ref] Serum N-Terminal Propeptide of Collagen Type I is Associated with the Number..., Leeming [/bib_ref].
Here we give an introduction and overview to the link between tumorigenesis, fibroblasts derived collagens and CAFs.
## Cancer associated fibroblasts -key players in cancer progression and desmoplasia
New insight into the role of CAFs have shown that these cells play a key-role in cancer progression. In the tumor microenvironment, transforming growth factor beta (TGF-β), platelet-derived growth factor and fibroblast growth factor-2, among others, secreted from malignant cells, attracts fibroblasts from neighboring tissue as well as aid in the transformation of normal fibroblast to CAFs within the tumor tissue [bib_ref] Cancer associated fibroblasts: the dark side of the coin, Cirri [/bib_ref] [bib_ref] Fibroblast heterogeneity in the cancer wound, Öhlund [/bib_ref] [bib_ref] TGF-β the master regulator of fibrosis. Pdf, Meng [/bib_ref] [bib_ref] Therapeutic effects of deleting cancer-associated fibroblasts in cholangiocarcinoma, Mertens [/bib_ref] [bib_ref] Platelet-derived growth factor-D and Rho GTPases regulate recruitment of cancer-associated fibroblasts in..., Cadamura [/bib_ref]. Up to 80% of the normal fibroblasts in breast tissue acquire the CAF phenotype during cancer progression [bib_ref] Fibroblasts in cancer, Kalluri [/bib_ref]. Interestingly, CAFs can also originate from epithelial cells, immune cells and endothelial cells [bib_ref] Discovery of endothelial to mesenchymal transition as a source for carcinoma-associated fibroblasts, Zeisberg [/bib_ref] emphasizing the complexity of this cell type. The CAF phenotype is characterized by changes in morphology and increased expression of myofibroblast markers such as alpha-smooth muscle actin (α-SMA), Vimentin, type XI collagen, fibronectin, fibroblast specific protein 1 (FSP-1) and fibroblast activating protein (FAP) [bib_ref] Carcinoma-associated fibroblasts: orchestrating the composition of malignancy, Gascard [/bib_ref]. Furthermore, CAFs show increased production of IM collagens [bib_ref] Aberrant type I and type III collagen gene expression in human breast..., Kauppila [/bib_ref] [bib_ref] A subset of myofibroblastic cancer-associated fibroblasts regulate collagen fiber elongation, which is..., Hanley [/bib_ref]. It is an ongoing discussion whether distinctive features between CAFs and myofibroblasts exist. The literature focusing on this topic is scarce and it seems that there is no consensus on what define quiescent fibroblasts, myofibroblasts and CAFs. Myofibroblasts (hepatic myofibroblasts) and CAFs (fibroblasts isolated from liver cancer patients) expressing α-SMA and Tenascin-C show similar apoptosis signaling compared to fibroblasts not expressing α-SMA and Tenascin-C (defined by the authors as quiescent fibroblasts) [bib_ref] Therapeutic effects of deleting cancer-associated fibroblasts in cholangiocarcinoma, Mertens [/bib_ref]. However, in another study performed by Öhlund and colleagues, it was shown that the transcriptional profiles between myofibroblasts (pancreatic stellate cells grown in monolayer) and CAFs (pancreatic stellate cells cocultured with tumor organoids) differs [bib_ref] Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer, Öhlund [/bib_ref]. Thus, the difficulties in defining what a CAF is and compare these cells/cell states [bib_ref] Cancer associated fibroblasts" -more than meets the eye, Madar [/bib_ref] to other cells is still an ongoing battle. To further complicate things, many studies have shown that different CAF subtypes exist based on differences in protein expression, paracrine signaling, tumorigenicity, invasion profile, ECM modifying capacities etc. [bib_ref] The biology and function of fibroblasts in cancer, Kalluri [/bib_ref] [bib_ref] Cancer-associated fibroblasts induce a collagen cross-link switch in tumor stroma, Pankova [/bib_ref] [bib_ref] Fibroblast heterogeneity in the cancer wound, Öhlund [/bib_ref] [bib_ref] Carcinoma-associated fibroblasts: orchestrating the composition of malignancy, Gascard [/bib_ref] [bib_ref] A subset of myofibroblastic cancer-associated fibroblasts regulate collagen fiber elongation, which is..., Hanley [/bib_ref] [bib_ref] Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer, Öhlund [/bib_ref] [bib_ref] Cancer: fibroblasts for all seasons, Wagner [/bib_ref] [bib_ref] Cancerassociated fibroblasts induce metalloprotease-independent cancer cell invasion of the basement membrane, Glentis [/bib_ref].
Several studies indicate that CAFs modulate epithelial transformation and promote cancer progression. As one example, CAFs have shown to initiate malignant transformation in non-malignant cells through overexpression of estrogen, TGF-β and hepatocyte growth factor [bib_ref] Breast stroma plays a dominant regulatory role in breast epithelial growth and..., Shekhar [/bib_ref] [bib_ref] Reconstruction of functionally normal and malignant human breast tissues in mice, Kuperwasser [/bib_ref]. As another example, a more mesenchymal phenotype has been observed for non-malignant prostate cells, when co-cultured with CAFs [bib_ref] Carcinoma-associated Fibroblasts Direct Tumor Progression of Initiated HumanProstatic Epithelium.pdf, Olumi [/bib_ref]. In addition The extracellular matrix during tumor progression. As the cancer cells invade the basement membrane (BM) the interstitial matrix (IM) becomes more and more desmoplastic characterized by an increased activity of cancer-associated fibroblasts (CAFs) and augmented volume of cross-linked type I, II, III, V, VI, XI, XXIV and XXVII collagens. In the later stages of tumor progression, desmoplasia pre-dominate the tumor microenvironment with signals from CAFs and IM collagens stimulating and sustaining the tumor progression to these in vitro examples, the initiation of cancer, by CAFs, have also been shown in vivo, where the injection of non-tumorigenic prostate cells, co-cultured with CAFs, lead to the formation of large tumors. On the contrary, no formation was evident when cells were cultured with normal fibroblasts [bib_ref] Breast stroma plays a dominant regulatory role in breast epithelial growth and..., Shekhar [/bib_ref]. Other than initiating cancer, CAFs have also been shown to sustain cancer progression and induce angiogenesis. Breast tumor tissue, isolated from mice, containing abundant amounts of CAFs shows increased vascularity compared to tissue with normal fibroblast [bib_ref] Stromal fibroblasts present in invasive human breast carcinomas promote tumor growth and..., Orimo [/bib_ref]. Glentis and colleagues, suggest that CAFs play a role in the invasion of cancer cells through the BM by pulling and stretching the BM resulting in small wholes which the cancer cells can squeeze through [bib_ref] Cancerassociated fibroblasts induce metalloprotease-independent cancer cell invasion of the basement membrane, Glentis [/bib_ref]. They also showed that especially invasive tumors from colon cancer patients are surrounded by a thick capsule of CAFs, which further suggests CAF involvement in invasion [bib_ref] Cancerassociated fibroblasts induce metalloprotease-independent cancer cell invasion of the basement membrane, Glentis [/bib_ref]. In line with this, CAFs also play a role in metastasis. Lung cancer cells treated with media from CAFs have increased migration potential compared to cells treated with media from normal fibroblasts [bib_ref] Cancer-associated fibroblasts enhance metastatic potential of lung cancer cells through IL-6/ STAT3..., Wang [/bib_ref]. This is further supported, by a study showing that cervical cancer cells co-transplanted with CAFs into mice leads to lymph node metastasis. In contrast, injections without CAFs do not lead to lymph node metastasis [bib_ref] Reconstitution of a metastatic-resistant tumor microenvironment with cancer-associated fibroblasts enables metastasis, Murata [/bib_ref]. The association between CAFs and lymph node metastasis has also been shown in esophageal squamous cell carcinoma in humans [bib_ref] Cancerassociated fibroblasts (CAFs) promote the lymph node metastasis of esophageal squamous cell..., Kashima [/bib_ref]. Several studies have also shown that CAFs play a role in inflammation by modulation of inflammatory components which promote tumor growth and metastasis [bib_ref] Distinct populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer, Öhlund [/bib_ref] [bib_ref] Cancer Associated Fibroblasts express pro-inflammatory factors in human breast and ovarian tumors, Erez [/bib_ref] [bib_ref] Inflammatory Role of Cancer-Associated Fibroblasts in Invasive Breast Tumors Revealed Using a..., Balachander [/bib_ref] (reviewed by [bib_ref] Inflammatory fibroblasts in cancer, Lim [/bib_ref]. Thus, these examples show the important role CAFs play in initiating and sustaining epithelial transformation and cancer progression across many different solid tumor types.
Aside from directly affecting cancer cells, CAFs are also major contributors to desmoplasia and remodeling of the ECM. Recent evidence indicates that CAFs modulate the desmoplastic reaction by affecting a wide variety of ECM proteins during tumorigenesis. A study has shown that CAFs take part in the assembly of fibronectin, which is highly abundant in the ECM and strongly involved in metastasis [bib_ref] Cancerassociated fibroblasts lead tumor invasion through integrin-beta3-dependent fibronectin assembly, Attieh [/bib_ref]. CAFs also express high amounts of the major ECM component hyaluronic acid, which has shown to encompass many structural and biological functions in tumor progression [bib_ref] Cancer-associated fibroblasts and the tumor microenvironment in malignant progression, Mccarthy [/bib_ref]. The oncogenes YAP/TAZ are suggested as being part of the remodeling processes exerted by CAFs. When the ECM becomes stiff, YAP/ TAZ gets transcriptionally active and promote CAF function which further stiffens the ECM (reviewed in [bib_ref] Mechanotransduction and YAP-dependent matrix remodelling is required for the generation and maintenance..., Calvo [/bib_ref]. The regulation of YAP/TAZ, resulting in CAF activity, is further regulated by the so called Rho family of small GTPases, which plays a role in CAF functioning and myofibroblast signaling [bib_ref] Platelet-derived growth factor-D and Rho GTPases regulate recruitment of cancer-associated fibroblasts in..., Cadamuro [/bib_ref] [bib_ref] Rac1, and Cdc42 differentially regulate SMA and collagen i expression in mesenchymal..., Ge [/bib_ref].
Some of the major steps in desmoplasia are cross-linking of collagens, fiber elongation and fiber realignment, which are associated with poor survival in cancer patients [bib_ref] A subset of myofibroblastic cancer-associated fibroblasts regulate collagen fiber elongation, which is..., Hanley [/bib_ref] [bib_ref] Highly aligned stromal collagen is a negative prognostic factor following pancreatic ductal..., Drifka [/bib_ref]. CAFs secrete increased amounts of MMPs and LOX-proteins, which catalyze these steps [bib_ref] Cancer-associated fibroblasts induce a collagen cross-link switch in tumor stroma, Pankova [/bib_ref] [bib_ref] A subset of myofibroblastic cancer-associated fibroblasts regulate collagen fiber elongation, which is..., Hanley [/bib_ref]. CAF secreted MMPs also play a key role in neovascularization because of the release of VEGF from degraded matrix [bib_ref] Basement membranes: structure, assembly and role in tumour angiogenesis, Kalluri [/bib_ref] [bib_ref] Cancerassociated fibroblasts induce metalloprotease-independent cancer cell invasion of the basement membrane, Glentis [/bib_ref] [bib_ref] Cancerassociated fibroblasts: their characteristics and their roles in tumor growth, Shiga [/bib_ref]. ECM proteins secreted and modulated by CAFs further recruit other cell types such as immune cells, which promote tumor progression [bib_ref] Cancer associated fibroblasts: the dark side of the coin, Cirri [/bib_ref] [bib_ref] Fibroblast heterogeneity in the cancer wound, Öhlund [/bib_ref]. Finally, a key step in desmoplasia, is the increased expression of fibroblast-derived collagens within the stroma. The accumulation of collagens, accompanied by increased cross-linking and stiffening of the tissue increase interstitial fluid pressure [bib_ref] High interstitial fluid pressure -an obstacle in cancer therapy, Heldin [/bib_ref]. This effect has been shown to reduce drug delivery of chemotherapy and immunotherapy [bib_ref] High interstitial fluid pressure -an obstacle in cancer therapy, Heldin [/bib_ref]. The stiffened tissue also play a role in tumor cell invasion, as the cross-linked collagens can create paths for the tumor cells to travel on [bib_ref] Collagen as a double-edged sword in tumor progression, Fang [/bib_ref].
Although consensus is that desmoplasia is a pro-tumorigenic event, results have emerged from mouse studies that have raised debate in the field. In one study, it has been shown that when the stromal content was reduced by deleting the sonic hedgehog protein in a pancreas cancer mouse model, the mice had more aggressive tumors as compared to control mice [bib_ref] Stromal elements act to restrain, rather than support, pancreatic ductal adenocarcinoma, Rhim [/bib_ref]. This was supported by similar findings, showing that the depletion of CAFs in mice led to much more aggressive tumors [bib_ref] Depletion of carcinoma-associated fibroblasts and fibrosis induces immunosuppression and accelerates pancreas cancer..., Özdemir [/bib_ref]. These findings do not exclude that desmoplasia is pro-tumorigenic, but suggest that a homeostatic restoration of the desmoplastic stroma, rather than its ablation, may be the best approach for eliminating tumor progression, as also suggested by Froeling and Kocher [bib_ref] Homeostatic restoration of desmoplastic stroma rather than its ablation slows pancreatic cancer..., Froeling [/bib_ref]. To further complicate matters, it has been suggested that some CAF subsets promote cancer, while others inhibit cancer [bib_ref] The biology and function of fibroblasts in cancer, Kalluri [/bib_ref] [bib_ref] Tumorpromoting Desmoplasia is disrupted by depleting FAP-expressing stromal cells, Lo [/bib_ref]. Albeit CAF biology and desmoplasia is complex, tumor tissue containing high amounts of CAFs have been reported to correlate with poor patient outcome in many different cancer types including colorectal, breast, tongue and esophageal cancer [bib_ref] Occurence of stromal myofibroblasts in the invasive ductal breast cancer tissue is..., Surowiak [/bib_ref] [bib_ref] Stromal features are predictive of disease mortality in oral cancer patients, Marsh [/bib_ref] [bib_ref] Myofibroblasts in the stroma of oral squamous cell carcinoma are associated with..., Kellermann [/bib_ref] [bib_ref] Stromal myofibroblasts predict disease recurrence for colorectal cancer, Tsujino [/bib_ref] [bib_ref] Stromal genes discriminate preinvasive from invasive disease, predict outcome, and highlight inflammatory..., Saadi [/bib_ref].
## Fibroblast derived stromal collagens and their contribution to tumorigenesis
While extensive research is currently going in the direction of CAF phenotype and their prognostic aspects, less in known regarding the collagens they produce. Are there functional differences in the collagen profile of tumors and does 'good' and 'bad' collagens exist in the tumor microenvironment as has been described for fibrosis [bib_ref] The good and the bad collagens of fibrosis -their role in signaling..., Karsdal [/bib_ref] , i.e. are collagen components originating from CAFs affecting tumor progression?
Collagens, and especially fibroblast-derived collagens (fibrillar collagens and the beaded filament type VI collagen), are extremely important in cancer. Most of these collagens are upregulated in cancer on both gene and protein level. They all modulate crucial steps in tumorigenesis such as proliferation, apoptosis, angiogenesis, invasion and metastasis. For many of the fibroblast-derived collagens specific chains of the collagens and pro-collagens have shown to be the effectors. Some studies even suggest that few of these collagens can inhibit tumorigenesis, and that different levels of collagens have different effects [bib_ref] Type III collagen directs stromal organization and limits metastasis in a murine..., Brisson [/bib_ref] [bib_ref] The effect of type III collagen on migration and invasion of human, Chintala [/bib_ref] [bib_ref] The spread of human lung cancer cells on collagens and its inhibition..., Hirai [/bib_ref]. This do suggest, that the turnover of fibroblast-collagens is important and relevant in the cancer setting and should be considered when exploring these collagens. Here we give an overview of these collagens and their contribution to tumorigenesis [fig_ref] Table 1: Overview of collagen type I, II, III, V, VI, XXIV and XXVII... [/fig_ref].
## Type i collagen
Type I collagen is the most abundant collagen throughout the body. It is the major component of the bone and is present in blood vessels, cornea, sclera, tendon, ligaments and skin. It is the most common collagen in the IM, where it has key structural roles. Apart from its structural role, type I collagen possess important growth factor binding potential, and via its binding to a variety of proteins regulate cell homeostasis [bib_ref] Type I collagen can function as a reservoir of basic fibroblast growth..., Kanematsu [/bib_ref].
A number of studies have shown that type I collagen play a significant role in bone related diseases, inclusive bone cancer and cancer-related bone metastasizes. Especially the turnover of type I collagen has shown to be important [bib_ref] Value of Ctelopeptide-cross-linked type I collagen, osteocalcin, bone-specific alkaline phosphatase and procollagen..., Zhao [/bib_ref] [bib_ref] Type I collagen receptor (alpha 2 beta 1) signaling promotes the growth..., Hall [/bib_ref] [bib_ref] N-Telopeptide of type I collagen Long-term dynamics in breast Cancer patients with..., Ferreira [/bib_ref].
Type I collagen is also dysregulated in other solid tumor types (than bone cancer) and can affect tumor cell behavior. Compared to healthy tissue, the amount of type I collagen is augmented in pancreas, colorectal, ovarian, breast and lung cancer [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Serum biomarkers reflecting specific tumor tissue remodeling processes are valuable diagnostic tools..., Willumsen [/bib_ref] [bib_ref] Excessive collagen turnover products are released during colorectal cancer progression and elevated..., Kehlet [/bib_ref] [bib_ref] Up-regulation of type I collagen during tumorigenesis of colorectal cancer revealed by..., Zou [/bib_ref].
Pancreas cancer cells exposed to type I collagen show increased proliferation, are less responsive to apoptosis, secrete higher amounts of TGF-β and show a strong reduction in E-cadherin expression [bib_ref] Type I collagen promotes the malignant phenotype of pancreatic ductal adenocarcinoma, Armstrong [/bib_ref] [bib_ref] Downregulation of E-cadherin gene expression by collagen type I and type III..., Menke [/bib_ref] [bib_ref] Type I collagen down-regulates E-cadherin expression by increasing PI3KCA in cancer cells, Cheng [/bib_ref]. Interestingly, Gao et al. found that tumor cells, in mouse breast tumor tissue, show high proliferative activity when located adjacent to type I collagen, whereas cells not in contact with type I collagen are quiescent [bib_ref] Multi-organ site metastatic reactivation mediated by non-canonical Discoidin domain receptor 1 signaling, Gao [/bib_ref].
Type I collagen has also been shown to affect metastasis, as exposure to type I collagen results in more invasive behavior in tumor cells [bib_ref] Downregulation of E-cadherin gene expression by collagen type I and type III..., Menke [/bib_ref]. In an in vivo breast cancer model, with accumulated type I collagen distribution, the amount of circulating tumor cells was increased compared to the amount in wild type mice. Moreover, the metastatic lesions were larger than in wild type [bib_ref] Elevated collagen-I augments tumor progressive signals, intravasation and metastasis of prolactin-induced estrogen..., Barcus [/bib_ref].
## Type ii collagen
Type II collagen is the main collagen in cartilage, where it constitutes 80% of the total collagen content. Within the joint, it provides stability and resiliency to stress. Forty percent of all bone cancers originates from cartilage, however bone cancers accounts for less than 0.2% of all cancersand therefore very little is known about type II collagen and its relation to cancer. However, a few studies have shown that type II collagen can affect cell behavior and that the type II collagen fragment PIIBNP can inhibit osteoclast survival and induce cell death in tumor cells [bib_ref] The type II collagen N-propeptide, PIIBNP, inhibits cell survival and bone resorption..., Hayashi [/bib_ref] [bib_ref] Citrullination of collagen II affects integrin-mediated cell adhesion in a receptor-specific manner, Sipilä [/bib_ref] [bib_ref] Type IIB procollagen NH2-propeptide induces death of tumor cells via interaction with..., Wang [/bib_ref].
## Type iii collagen
Type III collagen is the second most abundant collagen and is often distributed close to type I collagen. It is primarily found in vascular systems, intestine, liver, skin and lung. Like type I collagen, type III collagen distribution is augmented in many cancer diseases such as head and neck squamous cell cancer (HNSCC), breast, pancreas and colorectal cancer [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Extracellular matrix specific protein fingerprints measured in serum can seperate pancreatic cancer..., Willumsen [/bib_ref] [bib_ref] Aberrant type I and type III collagen gene expression in human breast..., Kauppila [/bib_ref] [bib_ref] Type I and III collagen degradation products in serum predict patient survival..., Nurmenniemi [/bib_ref] [bib_ref] Altered type III collagen turnover measured in pre-treatment serum predicts outcome in..., Jensen [/bib_ref]. In colon cancer, the distribution of type III collagen is especially augmented next to neovascular tissue [bib_ref] Aberrant type I and type III collagen gene expression in human breast..., Kauppila [/bib_ref].
Pancreas cancer cells grown on type III collagen show increased proliferation, migration and decreased expression of E-cadherin [bib_ref] Downregulation of E-cadherin gene expression by collagen type I and type III..., Menke [/bib_ref]. Moreover, type III collagen is involved in invasion and metastasis of glioblastoma cells. These cells show high invasion and migration response when exposed to type III collagen and antibodies against type III collagen inhibit these processes [bib_ref] The effect of type III collagen on migration and invasion of human, Chintala [/bib_ref]. Another study, report that collagen III is one of few genes that are modified, when invasive prostate cancer cells interact with bone marrow stromal cells, within the bone microenvironment. This interaction is crucial for the metastasis process, which further suggests an involvement of type III collagen in invasion and metastasis [bib_ref] The spread of human lung cancer cells on collagens and its inhibition..., Hirai [/bib_ref].
## Type v collagen
Type V collagen is a minor fibrillary collagen expressed in same tissues as collagen I and III, and helps in the formation of tissue specific matrices [bib_ref] Another look at collagen V and XI molecules, Fichard [/bib_ref]. Especially, the a3 chain of type V collagen has shown to be involved in cancer biology. When injecting breast tumor cells into mice deficient of the a3 chain in collagen 5 (Col5a3 −/− ) tumor growth is reduced and survival prolonged compared to wildtype littermates [bib_ref] α3 chains of type v collagen regulate breast tumour growth via glypican-1, Huang [/bib_ref]. In addition, Col5a3 −/− cancer cells injected into Col5a3 −/− and Col5a3 +/+ mice prolonged survival significantly in both genotypes compared to injection of cells containing the collagen V a3 chain [bib_ref] α3 chains of type v collagen regulate breast tumour growth via glypican-1, Huang [/bib_ref]. Thus, these two examples suggest that the presence of the collagen V a3 chain promote tumor growth.
## Type vi collagen
Type VI collagen is present in many tissues such as adipose, cartilage, skin, cornea, tendon, lung, skeletal muscle and dermis. It is located near the BM where it functions as a mediator between the BM and IM via its - Tissue distribution Main organic compound in bone. Also present in soft tissue.
[75]
- Tissue distribution in associated cancers
Major implications in bone cancer, and metastasis from bone to other solid tumors. Also described in breast, colorectal, ovarian, lung and pancreas cancer. [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Serum biomarkers reflecting specific tumor tissue remodeling processes are valuable diagnostic tools..., Willumsen [/bib_ref] [bib_ref] Excessive collagen turnover products are released during colorectal cancer progression and elevated..., Kehlet [/bib_ref] [bib_ref] Value of Ctelopeptide-cross-linked type I collagen, osteocalcin, bone-specific alkaline phosphatase and procollagen..., Zhao [/bib_ref] [bib_ref] Type I collagen receptor (alpha 2 beta 1) signaling promotes the growth..., Hall [/bib_ref] [bib_ref] N-Telopeptide of type I collagen Long-term dynamics in breast Cancer patients with..., Ferreira [/bib_ref] [bib_ref] Up-regulation of type I collagen during tumorigenesis of colorectal cancer revealed by..., Zou [/bib_ref] [bib_ref] Type I collagen promotes the malignant phenotype of pancreatic ductal adenocarcinoma, Armstrong [/bib_ref] [bib_ref] Downregulation of E-cadherin gene expression by collagen type I and type III..., Menke [/bib_ref] [bib_ref] Type I collagen down-regulates E-cadherin expression by increasing PI3KCA in cancer cells, Cheng [/bib_ref] [bib_ref] Multi-organ site metastatic reactivation mediated by non-canonical Discoidin domain receptor 1 signaling, Gao [/bib_ref] [bib_ref] Elevated collagen-I augments tumor progressive signals, intravasation and metastasis of prolactin-induced estrogen..., Barcus [/bib_ref] - Tumor promoting effects Associated with apoptosis, invasion, metastasis and proliferation. [bib_ref] Type I collagen promotes the malignant phenotype of pancreatic ductal adenocarcinoma, Armstrong [/bib_ref] [bib_ref] Downregulation of E-cadherin gene expression by collagen type I and type III..., Menke [/bib_ref] [bib_ref] Type I collagen down-regulates E-cadherin expression by increasing PI3KCA in cancer cells, Cheng [/bib_ref] [bib_ref] Multi-organ site metastatic reactivation mediated by non-canonical Discoidin domain receptor 1 signaling, Gao [/bib_ref] [bib_ref] Elevated collagen-I augments tumor progressive signals, intravasation and metastasis of prolactin-induced estrogen..., Barcus [/bib_ref] - Liquid biomarker potential Associated with bone metastasis in prostate and breast cancer patients. Increased in serum from colorectal, lung and pancreas cancer patients.
[ [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Extracellular matrix specific protein fingerprints measured in serum can seperate pancreatic cancer..., Willumsen [/bib_ref] [bib_ref] Serum biomarkers reflecting specific tumor tissue remodeling processes are valuable diagnostic tools..., Willumsen [/bib_ref] [bib_ref] Excessive collagen turnover products are released during colorectal cancer progression and elevated..., Kehlet [/bib_ref] [bib_ref] Serum N-Terminal Propeptide of Collagen Type I is Associated with the Number..., Leeming [/bib_ref] Collagen II
- Tissue distribution Main collagen in cartilage.- Tissue distribution in associated cancers
Associated with chondrosarcoma [bib_ref] The type II collagen N-propeptide, PIIBNP, inhibits cell survival and bone resorption..., Hayashi [/bib_ref] [bib_ref] Citrullination of collagen II affects integrin-mediated cell adhesion in a receptor-specific manner, Sipilä [/bib_ref] [bib_ref] Type IIB procollagen NH2-propeptide induces death of tumor cells via interaction with..., Wang [/bib_ref] - Tumor promoting effects Associated with cell death and survival [bib_ref] The type II collagen N-propeptide, PIIBNP, inhibits cell survival and bone resorption..., Hayashi [/bib_ref] [bib_ref] Citrullination of collagen II affects integrin-mediated cell adhesion in a receptor-specific manner, Sipilä [/bib_ref] [bib_ref] Type IIB procollagen NH2-propeptide induces death of tumor cells via interaction with..., Wang [/bib_ref] - Liquid biomarker potential n/a Collagen type III
- Tissue distribution Primarily found in the vascular system, intestine, liver, skin and lung.- Tissue distribution in associated cancers Implications in breast, colorectal, HNSCC and pancreas cancer. [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Extracellular matrix specific protein fingerprints measured in serum can seperate pancreatic cancer..., Willumsen [/bib_ref] [bib_ref] Aberrant type I and type III collagen gene expression in human breast..., Kauppila [/bib_ref] [bib_ref] Type I and III collagen degradation products in serum predict patient survival..., Nurmenniemi [/bib_ref] [bib_ref] Altered type III collagen turnover measured in pre-treatment serum predicts outcome in..., Jensen [/bib_ref] - Tumor promoting effects Associated with invasion, metastasis, migration and proliferation. [bib_ref] The effect of type III collagen on migration and invasion of human, Chintala [/bib_ref] [bib_ref] The spread of human lung cancer cells on collagens and its inhibition..., Hirai [/bib_ref] [bib_ref] Downregulation of E-cadherin gene expression by collagen type I and type III..., Menke [/bib_ref] - Liquid biomarker potential Augmented in serum from ovarian, breast, colorectal, melanoma and pancreas cancer patients. Able to predict pancreas cancer patients most likely to respond to treatment. [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Excessive collagen turnover products are released during colorectal cancer progression and elevated..., Kehlet [/bib_ref] [bib_ref] Altered type III collagen turnover measured in pre-treatment serum predicts outcome in..., Jensen [/bib_ref] [bib_ref] Extracellular matrix (ECM) circulating peptide biomarkers as potential predictors of survival in..., Wang [/bib_ref] Collagen type V
- Tissue distribution Primarily in same tissues as collagen type I and III. [bib_ref] Another look at collagen V and XI molecules, Fichard [/bib_ref] - Tissue distribution in associated cancers
Associated with breast cancer. [bib_ref] α3 chains of type v collagen regulate breast tumour growth via glypican-1, Huang [/bib_ref] - Tumor promoting effects Associated with tumor growth. [bib_ref] α3 chains of type v collagen regulate breast tumour growth via glypican-1, Huang [/bib_ref] - Liquid biomarker potential n/a Collagen type VI
- Tissue distribution Present in many tissues such as adipose, cartilage, skin, cornea, tendon, lung, skeletal muscle and dermis.
[97]
- Tissue distribution in associated cancers Described in breast, colorectal, ovarian, gliomas, melanomas and pancreas cancer. [bib_ref] Collagen VI in cancer and its biological mechanisms, Chen [/bib_ref] - Tumor promoting effects Associated with apoptosis, drug resistance, inflammation, invasion, metastasis and proliferation. [bib_ref] Collagen VI at a glance, Cescon [/bib_ref] [bib_ref] Collagen VI in cancer and its biological mechanisms, Chen [/bib_ref] [bib_ref] Endotrophin -linking obesity with aggressive tumor growth, Park [/bib_ref] [bib_ref] Production of type VI collagen by human macrophages: a new dimension in..., Schnoor [/bib_ref] [bib_ref] The differential effect of endothelial cell factors on in vitro motility of..., Wright [/bib_ref] [bib_ref] Remodeling of the extracellular matrix through overexpression of collagen VI contributes to..., Sherman-Baust [/bib_ref] [bib_ref] Gene expression profiling of a clonal isolate of oxaliplatin-resistant ovarian carcinoma cell..., Varma [/bib_ref] [bib_ref] Inhibition of endotrophin, a cleavage product of collagen VI, confers cisplatin sensitivity..., Park [/bib_ref] [bib_ref] Adipocyte-derived collagen VI affects early mammary tumor progression in vivo, demonstrating a..., Iyengar [/bib_ref] - Liquid biomarker potential Augmented in serum from melanoma and pancreas cancer patients. [bib_ref] Laminin, hyaluronan, tenascin-C and type VI collagen levels in sera from patients..., Burchardt [/bib_ref] [bib_ref] Clinical significance of serum COL6A3 in pancreatic ductal adenocarcinoma, Kang [/bib_ref] Collagen type XI
- Tissue distribution Distributed in low levels in skeletal muscle, trabecular bone, tendons, testis, trachea, articular cartilage, lung, placenta and brain.
[ [bib_ref] Collagen type XI α1 facilitates head and neck squamous cell cancer growth..., Sok [/bib_ref] [bib_ref] A gene expression signature that defines breast cancer metastases, Ellsworth [/bib_ref] [bib_ref] A potential role of collagens expression in distinguishing between premalignant and malignant..., Zhao [/bib_ref] [bib_ref] Overexpression of COL11A1 by Cancerassociated fibroblasts: clinical relevance of a stromal marker..., García-Pravia [/bib_ref] [bib_ref] Great potential of a panel of multiple hMTH1, SPD, ITGA11 and COL11A1..., Chong [/bib_ref] [bib_ref] Colorectal carcinogenesis is associated with stromal expression of COL11A1 and COL5A2, Fischer [/bib_ref] [bib_ref] Combined gene expression analysis of whole-tissue and microdissected pancreatic ductal adenocarcinoma identifies..., Badea [/bib_ref] - Tissue distribution in associated cancers
Extremely augmented in colorectal and HNSCC cancer. Also associated with breast, gastric, lung, ovarian and pancreas cancer.
[ [bib_ref] Collagen type XI α1 facilitates head and neck squamous cell cancer growth..., Sok [/bib_ref] [bib_ref] A gene expression signature that defines breast cancer metastases, Ellsworth [/bib_ref] [bib_ref] A potential role of collagens expression in distinguishing between premalignant and malignant..., Zhao [/bib_ref] [bib_ref] Overexpression of COL11A1 by Cancerassociated fibroblasts: clinical relevance of a stromal marker..., García-Pravia [/bib_ref] [bib_ref] Great potential of a panel of multiple hMTH1, SPD, ITGA11 and COL11A1..., Chong [/bib_ref] [bib_ref] Colorectal carcinogenesis is associated with stromal expression of COL11A1 and COL5A2, Fischer [/bib_ref] [bib_ref] Combined gene expression analysis of whole-tissue and microdissected pancreatic ductal adenocarcinoma identifies..., Badea [/bib_ref] - Tumor promoting effects Highly implicated in CAF biology. Also associated with invasion, metastasis and proliferation. [bib_ref] Collagen type XI α1 facilitates head and neck squamous cell cancer growth..., Sok [/bib_ref] [bib_ref] A gene expression signature that defines breast cancer metastases, Ellsworth [/bib_ref] [bib_ref] A potential role of collagens expression in distinguishing between premalignant and malignant..., Zhao [/bib_ref] [bib_ref] Overexpression of COL11A1 by Cancerassociated fibroblasts: clinical relevance of a stromal marker..., García-Pravia [/bib_ref] [bib_ref] Great potential of a panel of multiple hMTH1, SPD, ITGA11 and COL11A1..., Chong [/bib_ref] [bib_ref] Colorectal carcinogenesis is associated with stromal expression of COL11A1 and COL5A2, Fischer [/bib_ref] [bib_ref] Combined gene expression analysis of whole-tissue and microdissected pancreatic ductal adenocarcinoma identifies..., Badea [/bib_ref] [bib_ref] COL11A1 promotes tumor progression and predicts poor clinical outcome in ovarian cancer, Wu [/bib_ref] [bib_ref] Multi-cancer computational analysis reveals invasion-associated variant of desmoplastic reaction involving INHBA, THBS2..., Kim [/bib_ref] [bib_ref] Identification of candidate circulating cisplatin-resistant biomarkers from epithelial ovarian carcinoma cell secretomes, Teng [/bib_ref] [bib_ref] Prognostic gene-expression signature of carcinoma-associated fi broblasts in non-small cell lung cancer, Navab [/bib_ref] [bib_ref] A collagenremodeling gene signature regulated by TGF-β signaling is associated with metastasis..., Cheon [/bib_ref] - Liquid biomarker potential n/a Collagen type XXIV
- Tissue distribution Distributed in ovaries, testis, liver, spleen, kidney, muscle and bone. [bib_ref] The collagen superfamily: from the extracellular matrix to the cell membrane, Ricard-Blum [/bib_ref] [bib_ref] Collagen XXIV , a Vertebrate Fibrillar Collagen with Structural Features of Invertebrate..., Koch [/bib_ref] [bib_ref] Collagen XXIV (Col24α1) promotes osteoblastic differentiation and mineralization.Pdf, Wang [/bib_ref] - Tissue distribution in associated cancers
Associated with HNSCC.- Tumor promoting effects Associated with cell differentiation.- Liquid biomarker potential n/a Collagen type XXVII many binding sites in both matrices. It can bind to a wide variety of proteins such as type I, II, IV, XIV collagen, integrin's, fibronectin, tenascin etc. Type VI collagen has many roles covering structural purposes to more cell-specific functions including regulation of apoptosis, proliferation, differentiation and maintenance of cell stemness [bib_ref] Collagen VI at a glance, Cescon [/bib_ref]. Collagen VI expression is increased in many human tumors such as glioblastomas, melanomas, ovarian, pancreatic, breast and colon cancer [bib_ref] Collagen VI in cancer and its biological mechanisms, Chen [/bib_ref]. In vitro and in vivo studies have shown that collagen VI increase proliferation and decrease apoptosis in breast, melanoma and glioblastoma cell lines [bib_ref] Collagen VI at a glance, Cescon [/bib_ref] [bib_ref] Collagen VI in cancer and its biological mechanisms, Chen [/bib_ref]. Apart from its direct stimulatory effects on tumor cells, collagen VI also affects the tumor microenvironment by promoting angiogenesis and inflammation [bib_ref] Collagen VI in cancer and its biological mechanisms, Chen [/bib_ref] [bib_ref] Endotrophin -linking obesity with aggressive tumor growth, Park [/bib_ref]. Collagen VI deficiency (col6 −/− ) inhibit endothelial cell growth and sprouting of new vessels in a melanoma mouse model. Regarding inflammation, macrophages has been shown to produce type VI collagen, which in this context, modulate cell-to-matrix and cell-to-cell interactions [bib_ref] Production of type VI collagen by human macrophages: a new dimension in..., Schnoor [/bib_ref]. Lastly, type VI collagen has shown to affect the invasion-profile of glioblastoma and lung-cancer cell [bib_ref] The differential effect of endothelial cell factors on in vitro motility of..., Wright [/bib_ref]. A number of studies have shown, that the a3 chain and the C5 domain of the a3 chain, also called endotrophin is involved in many hallmarks in cancer such as promoting proliferation, angiogenesis, metastasis and chemotherapy resistance. Type VI collagen a3 is distributed in high amounts in lung, ovarian, pancreatic, colon and breast cancer tissues [bib_ref] Collagen VI in cancer and its biological mechanisms, Chen [/bib_ref]. Endotrophin has been found to promote metastasis in breast cancer and recruit endothelial cells to the tumor microenvironment [bib_ref] Endotrophin -linking obesity with aggressive tumor growth, Park [/bib_ref]. This study also reported that endotrophin facilitate tumor cell proliferation and metastasis through TGF-β activation as well as promote inflammation in the tumor microenvironment by upregulating inflammatory markers such as interleukin-6 and TNF-a [bib_ref] Endotrophin -linking obesity with aggressive tumor growth, Park [/bib_ref]. In the context of chemotherapy resistance collagen VI a3 is one of the most highly expressed genes in cisplatin and oxaliplatin resistant ovarian cancer cells [bib_ref] Remodeling of the extracellular matrix through overexpression of collagen VI contributes to..., Sherman-Baust [/bib_ref] [bib_ref] Gene expression profiling of a clonal isolate of oxaliplatin-resistant ovarian carcinoma cell..., Varma [/bib_ref]. In addition, endotrophin is highly upregulated in cisplatin resistant breast tumor cells, and inhibition of endotrophin lead to cisplatin sensitivity in a breast tumor mouse model [bib_ref] Inhibition of endotrophin, a cleavage product of collagen VI, confers cisplatin sensitivity..., Park [/bib_ref]. Metallothioneins, which are associated with cisplatin resistance, are highly upregulated in breast cancer cells treated with collagen VI, which could be one of the explanations for the chemotherapy resistance, as suggested by Iangyar et al. [bib_ref] Adipocyte-derived collagen VI affects early mammary tumor progression in vivo, demonstrating a..., Iyengar [/bib_ref].
## Type xi collagen
Type XI collagen is present in low levels in skeletal muscle, trabecular bone, tendons, testis, trachea, articular cartilage, lung, placenta and in the brain. It is a minor fibrillar collagen, which co-polymerize with type II collagen and type IX collagen. In cartilage, it is extremely important for proper function, as absence of type XI collagen lead to abnormal thickening of the tissue. Collagen XI has long been suspected to be of high impact in cancer formation, and especially the a1 chain of collagen XI has shown to be an important player in various cancer diseases. The gene signature of type XI collagen is upregulated in breast, gastric, pancreatic, and non-small lung cancer. Interestingly, in both colon and HNSCC the expression is extremely increased with almost no expression in healthy controls [bib_ref] Collagen type XI α1 facilitates head and neck squamous cell cancer growth..., Sok [/bib_ref] [bib_ref] A gene expression signature that defines breast cancer metastases, Ellsworth [/bib_ref] [bib_ref] A potential role of collagens expression in distinguishing between premalignant and malignant..., Zhao [/bib_ref] [bib_ref] Overexpression of COL11A1 by Cancerassociated fibroblasts: clinical relevance of a stromal marker..., García-Pravia [/bib_ref] [bib_ref] Great potential of a panel of multiple hMTH1, SPD, ITGA11 and COL11A1..., Chong [/bib_ref] [bib_ref] Colorectal carcinogenesis is associated with stromal expression of COL11A1 and COL5A2, Fischer [/bib_ref] [bib_ref] Combined gene expression analysis of whole-tissue and microdissected pancreatic ductal adenocarcinoma identifies..., Badea [/bib_ref]. Knock down of type XIa1 collagen in HNSCC and ovarian cancer cell lines, significantly decrease proliferation, invasion and migration compared to controls, which highlight type XI collagens importance in cancer [bib_ref] Collagen type XI α1 facilitates head and neck squamous cell cancer growth..., Sok [/bib_ref] [bib_ref] COL11A1 promotes tumor progression and predicts poor clinical outcome in ovarian cancer, Wu [/bib_ref]. In breast and ovarian cancer collagen XIa1 has also been associated with resistant to chemotherapy [bib_ref] Multi-cancer computational analysis reveals invasion-associated variant of desmoplastic reaction involving INHBA, THBS2..., Kim [/bib_ref] [bib_ref] Identification of candidate circulating cisplatin-resistant biomarkers from epithelial ovarian carcinoma cell secretomes, Teng [/bib_ref].
Type XI collagen is highly associated with CAFs. CAFs originating from HNSCC, lung cancer and pancreas cancer tissue express higher levels of collagen XIa1 than cells arrived from healthy tissue [bib_ref] Collagen type XI α1 facilitates head and neck squamous cell cancer growth..., Sok [/bib_ref] [bib_ref] Overexpression of COL11A1 by Cancerassociated fibroblasts: clinical relevance of a stromal marker..., García-Pravia [/bib_ref] [bib_ref] Prognostic gene-expression signature of carcinoma-associated fi broblasts in non-small cell lung cancer, Navab [/bib_ref]. In ovarian and pancreatic cancer CAFs strongly stain for collagen XIa1, compared to no staining in epithelial cancer cells and healthy tissue [bib_ref] Overexpression of COL11A1 by Cancerassociated fibroblasts: clinical relevance of a stromal marker..., García-Pravia [/bib_ref] [bib_ref] A collagenremodeling gene signature regulated by TGF-β signaling is associated with metastasis..., Cheon [/bib_ref].
## Type xxiv
Type XXIV collagen is expressed in ovaries, testis, liver, spleen, lung, kidney, muscle and bone and is located close to type I and V collagen [bib_ref] The collagen superfamily: from the extracellular matrix to the cell membrane, Ricard-Blum [/bib_ref] [bib_ref] Collagen XXIV , a Vertebrate Fibrillar Collagen with Structural Features of Invertebrate..., Koch [/bib_ref] [bib_ref] Collagen XXIV (Col24α1) promotes osteoblastic differentiation and mineralization.Pdf, Wang [/bib_ref].
As with type II collagen very little is known regarding type XXIV collagen in relation to cancer. Type XXIV collagen has been associated with osteoblast differentiation [fig_ref] Table 1: Overview of collagen type I, II, III, V, VI, XXIV and XXVII... [/fig_ref] Overview of collagen type I, II, III, V, VI, XXIV and XXVII and their distribution in healthy tissue, cancer tissue, tumor promoting effects and liquid biomarker potential (Continued)
## Collagen type description reference
- Tissue distribution Expressed in the developing eyes, ears, lungs, heart and arteries. [bib_ref] Identification, characterization and expression analysis of a new fibrillar collagen gene, Pace [/bib_ref] [bib_ref] Collagen XXVII is developmentally regulated and forms thin fibrillar structures distinct from..., Plumb [/bib_ref] - Tissue distribution in associated cancers n/a
- Tumor promoting effects n/a
- Liquid biomarker potential n/a HNSCC Head and neck squamous cell carcinoma, n/a not available with the expression increased in tumor tissue from patients suffering from HNSCC.
## Type xxvii
Like type XXIV collagen, type XXVII is a relatively poorly characterized collagen. During embryogenesis in mice COL27A is expressed in the developing eyes, ears, lungs, heart and arteries [bib_ref] Identification, characterization and expression analysis of a new fibrillar collagen gene, Pace [/bib_ref] [bib_ref] Collagen XXVII is developmentally regulated and forms thin fibrillar structures distinct from..., Plumb [/bib_ref]. However, in adults it is primarily expressed in cartilage, and is therefore thought to play a role in the development phases. Type XXVII collagens role in cancer is yet to be investigated.
## Stromal derived biomarkers in clinical cancer research
A number of studies have investigated the possibility of using CAFs as prognostic markers in different cancer diseases. The most widely CAF biomarkers used for this are a-SMA, Vimentin, collagen XIa, fibronectin, FSP-1 and FAP. In esophageal cancer a-SMA and FSP-1 positive staining correlates with larger tumor size, advanced T-stage and shorter survival [bib_ref] The Prognostic Significance of Cancer-Associated Fibroblasts in Esophageal Squamous Cell Carcinoma, Ha [/bib_ref]. FAP is highly expressed in CAFs and present in many different cancer types, and has been associated with shorter survival in lung, esophageal and breast cancer [bib_ref] Cancerassociated fibroblasts (CAFs) promote the lymph node metastasis of esophageal squamous cell..., Kashima [/bib_ref] [bib_ref] Prognostic relevance of cancer-associated fibroblasts in human cancer, Paulsson [/bib_ref]. CAFs are very complex cells and the CAF markers used today display cellular overlaps, and have to be used in combinations [bib_ref] A peek into cancer-associated fibroblasts: origins, functions and translational impact, Lebleu [/bib_ref]. Therefore, developing specific CAF biomarkers or biomarkers measuring CAF activity, i.e. disease progression, should be of high priority. The existing CAF biomarkers are mainly based on immunohistochemistry, which rely on tissue biopsies. Although such tissue biomarkers are still the golden standard for tumor characterization, there are several benefits of developing biomarkers based on liquid biopsies (e.g. serum, plasma, urine). Besides being non-invasive, cost-effective and highly repeatable, liquid biopsies are also a real-time representative for the entire tumor heterogeneity, and not just a snapshot of the tumor tissue here and now.
The formation and degradation of fibroblast-derived collagens, during desmoplasia, are mediated by CAFs [bib_ref] Carcinoma-associated fibroblasts: orchestrating the composition of malignancy, Gascard [/bib_ref]. Thus, collagen fragments could be a measure of CAF activity. Interestingly, formation-and degradation products, in serum, from fibroblast-derived collagens show diagnostic and prognostic value. Degradation products from collagen I are significantly increased in colorectal cancer and able to differentiate stage IV colorectal cancer from stage I-III. [bib_ref] Excessive collagen turnover products are released during colorectal cancer progression and elevated..., Kehlet [/bib_ref]. The same trend is seen in ovarian, breast, lung and pancreas cancer patients, where degradation products from collagen I can distinguish cancer patients from healthy controls [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref] [bib_ref] Extracellular matrix specific protein fingerprints measured in serum can seperate pancreatic cancer..., Willumsen [/bib_ref] [bib_ref] Serum biomarkers reflecting specific tumor tissue remodeling processes are valuable diagnostic tools..., Willumsen [/bib_ref]. Moreover, a strong association between formation products from collagen I and the amount of bone metastasizes is seen in prostate and breast cancer [bib_ref] Serum N-Terminal Propeptide of Collagen Type I is Associated with the Number..., Leeming [/bib_ref].
Collagen III formation and degradation products are elevated in ovarian and breast cancer patients, and capable of distinguishing cancer patients from healthy controls [bib_ref] Collagen degradation products measured in serum can separate ovarian and breast cancer..., Bager [/bib_ref]. This is also shown for colorectal cancer where collagen III products are significantly elevated and correlate with tumor stage [bib_ref] Excessive collagen turnover products are released during colorectal cancer progression and elevated..., Kehlet [/bib_ref]. Interestingly, the ratio of formation and degradation markers of collagen III has shown to be capable of predicting pancreas patients most likely to respond to the hyaluronan targeting drug PEGPH20 (pegvorhyaluronidase alfa) [bib_ref] Extracellular matrix (ECM) circulating peptide biomarkers as potential predictors of survival in..., Wang [/bib_ref]. In addition, a high ratio predicts increased overall survival in melanoma patients [bib_ref] Altered type III collagen turnover measured in pre-treatment serum predicts outcome in..., Jensen [/bib_ref]. Lastly, serum levels of collagen VI are increased in melanoma and pancreatic cancer patients [bib_ref] Laminin, hyaluronan, tenascin-C and type VI collagen levels in sera from patients..., Burchardt [/bib_ref] [bib_ref] Clinical significance of serum COL6A3 in pancreatic ductal adenocarcinoma, Kang [/bib_ref].
Another potential role of collagen biomarkers are related to anti-TGF-β therapies that are emerging as novel treatments options, in particular in the immuno-oncology setting. TGF-β is a complex molecule with many roles in cancer [bib_ref] Remodeling of the extracellular matrix through overexpression of collagen VI contributes to..., Sherman-Baust [/bib_ref] [bib_ref] TGFβ the molecular Jekyll and Hyde.Pdf, Bierie [/bib_ref] amongst others TGF-β stimulate CAFs to produce collagens [bib_ref] TGF-β the master regulator of fibrosis. Pdf, Meng [/bib_ref] [bib_ref] Cytokine mediated tissue fibrosis, Borthwicka [/bib_ref]. Hence collagen turnover fragments may be predictive of a TGF-β driven phenotype and hence be used to identify patients benefiting from such treatment. In addition, these collagen biomarkers may be used to monitor on target effects of TGF-β and reveal valued information on mode of action of the compound investigated. A recent study has shown that the assembly of collagens can trap T-cells preventing them to access the tumor, and induce T-cell dependent cell death [bib_ref] TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of..., Mariathasan [/bib_ref]. This complicate the use of immune therapy and could be a reason to why only a subset of patients respond to therapy. In the last mentioned study, the occurrence of TGF-β producing fibroblasts was strongly associated with lack of therapy response [bib_ref] TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of..., Mariathasan [/bib_ref]. In this respect, collagen levels have the potential to be used as precision medicine to select patients most likely to respond to treatment.
# Conclusion
Alterations in tissue microarchitecture is a hallmark of cancer driven by CAFs and the associated deposition of collagens in the tumor stroma, which amongst other things leads to desmoplasia, poor prognosis and therapy resistance. In this review we have highlighted the link between CAFs, the fibrillar collagens produced by CAFs, and tumorigenesis. We provide a rationale for studying CAF-derived collagens in greater detail, to improve the understanding of tumor biology and patient characteristics. Lastly, we argue that a major biomarker potential lies in the fact that these collagen products can be measured in a liquid biopsy, providing a surrogate measure of desmoplasia and CAF activity. Future biomarker research should focus on implementing such biomarker tools in the clinical setting for phenotyping of cancer patients and potentially for predicting and monitoring response to treatment.
Abbreviations BM: Basement membrane; CAF: Cancer associated fibroblast; ECM: Extracellular matrix; FAP: Fibroblast activating protein; FSP-1: Fibroblast specific protein 1; HNSCC: Head and neck squamous cell cancer; IM: Interstitial matrix; LOX: Lysyl oxidase; MMP: Metalloproteinases; n/a: Not available; TGF-β : Transforming growth factor beta; α-SMA: alpha-smooth muscle actin
[fig] Figure 1: Collagens within the basement membrane and interstitial matrix. Schematic drawing of the structure and localization of network-forming collagens (type IV collagen), beaded filament (type VI collagen) and fibril-forming collagens (type I, II, III, V, XI, XXIV and XXVII collagens) [/fig]
[table] Table 1: Overview of collagen type I, II, III, V, VI, XXIV and XXVII and their distribution in healthy tissue, cancer tissue, tumor promoting effects and liquid biomarker potential [/table]
|
Deep Phenotyping of Chinese Electronic Health Records by Recognizing Linguistic Patterns of Phenotypic Narratives With a Sequence Motif Discovery Tool: Algorithm Development and Validation
Background: Phenotype information in electronic health records (EHRs) is mainly recorded in unstructured free text, which cannot be directly used for clinical research. EHR-based deep-phenotyping methods can structure phenotype information in EHRs with high fidelity, making it the focus of medical informatics. However, developing a deep-phenotyping method for non-English EHRs (ie, Chinese EHRs) is challenging. Although numerous EHR resources exist in China, fine-grained annotation data that are suitable for developing deep-phenotyping methods are limited. It is challenging to develop a deep-phenotyping method for Chinese EHRs in such a low-resource scenario.Objective: In this study, we aimed to develop a deep-phenotyping method with good generalization ability for Chinese EHRs based on limited fine-grained annotation data.Methods:The core of the methodology was to identify linguistic patterns of phenotype descriptions in Chinese EHRs with a sequence motif discovery tool and perform deep phenotyping of Chinese EHRs by recognizing linguistic patterns in free text. Specifically, 1000 Chinese EHRs were manually annotated based on a fine-grained information model, PhenoSSU (Semantic Structured Unit of Phenotypes). The annotation data set was randomly divided into a training set (n=700, 70%) and a testing set (n=300, 30%). The process for mining linguistic patterns was divided into three steps. First, free text in the training set was encoded as single-letter sequences (P: phenotype, A: attribute). Second, a biological sequence analysis tool-MEME (Multiple Expectation Maximums for Motif Elicitation)-was used to identify motifs in the single-letter sequences. Finally, the identified motifs were reduced to a series of regular expressions representing linguistic patterns of PhenoSSU instances in Chinese EHRs. Based on the discovered linguistic patterns, we developed a deep-phenotyping method for Chinese EHRs, including a deep learning-based method for named entity recognition and a pattern recognition-based method for attribute prediction.Results:In total, 51 sequence motifs with statistical significance were mined from 700 Chinese EHRs in the training set and were combined into six regular expressions. It was found that these six regular expressions could be learned from a mean of 134 (SD 9.7) annotated EHRs in the training set. The deep-phenotyping algorithm for Chinese EHRs could recognize PhenoSSU instances with an overall accuracy of 0.844 on the test set. For the subtask of entity recognition, the algorithm achieved an F1 score of 0.898 with the Bidirectional Encoder Representations from Transformers-bidirectional long short-term memory and J Med Internet Res 2022 | vol. 24 | iss. 6 | e37213 | p. 1 https://www.jmir.org/2022/6/e37213 (page number not for citation purposes) Li et al JOURNAL OF MEDICAL INTERNET RESEARCHXSL - FORenderX conditional random field model; for the subtask of attribute prediction, the algorithm achieved a weighted accuracy of 0.940 with the linguistic pattern-based method.Conclusions:We developed a simple but effective strategy to perform deep phenotyping of Chinese EHRs with limited fine-grained annotation data. Our work will promote the second use of Chinese EHRs and give inspiration to other non-English-speaking countries.Res 2022;24(6):e37213)(J Med Internet
# Introduction
Currently, electronic health records (EHRs) are increasingly becoming an important source for clinical data mining and analysis [bib_ref] VA Million Veteran Program. Clinical knowledge extraction via sparse embedding regression (KESER)..., Hong [/bib_ref]. Phenotype information that describes patients' clinical manifestations is one of the most valuable clinical information types in EHRs [bib_ref] Time to integrate clinical and research informatics, Katzan [/bib_ref]. However, phenotype information in EHRs is mainly recorded in free text, which computers have difficulty using directly [bib_ref] Using artificial intelligence with natural language processing to combine electronic health record's..., Elkin [/bib_ref] [bib_ref] Clinical information extraction applications: A literature review, Wang [/bib_ref]. Therefore, it is important to develop natural language processing (NLP) technology to effectively structure phenotype information in free text. The NLP technology for structuring phenotype information in EHRs is called EHR-based phenotyping [bib_ref] EHR-based phenotyping: Bulk learning and evaluation, Chiu [/bib_ref].
There are two key factors involved in EHR-based phenotyping [bib_ref] Deep EHR: A survey of recent advances in deep learning techniques for..., Shickel [/bib_ref]. The first factor is the development of an information model that can define the normalized target of phenotyping [bib_ref] Clinical element models in the SHARPn consortium, Oniki [/bib_ref]. The second factor is the development of a phenotyping algorithm that can process phenotype information into a predefined information model [bib_ref] Extracting entities with attributes in clinical text via joint deep learning, Shi [/bib_ref]. In recent years, the focus of EHR-based phenotyping methods has shifted from the coarse-grained level to the fine-grained level [bib_ref] High-fidelity phenotyping: Richness and freedom from bias, Hripcsak [/bib_ref] [bib_ref] Semantic integration of clinical laboratory tests from electronic health records for deep..., Zhang [/bib_ref]. Compared with coarse-grained phenotyping, fine-grained phenotyping can capture more phenotype details, including the phenotype concept and its associated attributes [bib_ref] Deep phenotyping: The details of disease, Delude [/bib_ref]. For example, in the free-text description "a sudden severe pain in the right-lower abdomen," a fine-grained deep-phenotyping method not only considers the phenotype "pain" but also its associated attributes of body location ("abdomen"), temporal pattern ("acute"), and severity ("severe"). EHR-based phenotyping that can characterize phenotype details at a fine-grained level is called EHR-based deep phenotyping [bib_ref] Deep phenotyping: Embracing complexity and temporality-Towards scalability, portability, and interoperability, Weng [/bib_ref].
Deep-phenotyping methods can characterize phenotype information in a high-fidelity way, which can potentially improve the accuracy of EHR-based applications, such as disease diagnosis and treatment [bib_ref] Deep phenotyping for precision medicine, Robinson [/bib_ref]. Hence, deep phenotyping has become the focus of medical informatics. In recent years, a series of deep-phenotyping methods for English EHRs have been developed. For example, Peterson et al [bib_ref] A corpus-driven standardization framework for encoding clinical problems with HL7 FHIR, Peterson [/bib_ref] used the MetaMap tool [bib_ref] An overview of MetaMap: Historical perspective and recent advances, Aronson [/bib_ref] to recognize phenotype concepts in EHRs, along with a neural network model to predict attribute values associated with phenotypes. They finally characterized English EHRs with the Fast Healthcare Interoperability Resources (FHIR) model [bib_ref] The Fast Health Interoperability Resources (FHIR) standard: Systematic literature review of implementations,..., Ayaz [/bib_ref]. Xu et al [bib_ref] Applying a deep learning-based sequence labeling approach to detect attributes of medical..., Xu [/bib_ref] developed a bidirectional long short-term memory and conditional random field (Bi-LSTM-CRF) model to recognize phenotype concepts in EHRs, together with a machine learning method to predict attribute values, and finally represented the phenotype information in English EHRs with the clinical element model (CEM) [bib_ref] A semantic-web oriented representation of the clinical element model for secondary use..., Tao [/bib_ref]. Compared to the progress of deep-phenotyping English EHRs, the method for deep-phenotyping Chinese EHRs is still in its infancy. Regarding the existence of linguistic differences, the established strategies [bib_ref] A corpus-driven standardization framework for encoding clinical problems with HL7 FHIR, Peterson [/bib_ref] [bib_ref] Applying a deep learning-based sequence labeling approach to detect attributes of medical..., Xu [/bib_ref] [bib_ref] Natural language processing and machine learning methods to characterize unstructured patient-reported outcomes:..., Lu [/bib_ref] [bib_ref] Artificial intelligence-based prediction of lung cancer risk using nonimaging electronic medical records:..., Yeh [/bib_ref] for deep-phenotyping English EHRs cannot be directly used for Chinese EHRs. Moreover, developing a deep-phenotyping algorithm requires fine-grained annotation data. However, it is hard to obtain a large volume of annotation data because of the high annotation cost. This means that the development of a deep-phenotyping algorithm for Chinese EHRs suffers from the challenge of low-resource scenarios [bib_ref] Extracting entities with attributes in clinical text via joint deep learning, Shi [/bib_ref] , so it is worth considering how to develop a generalized algorithm for deep-phenotyping Chinese EHRs with limited fine-grained annotated data.
In previous work, we developed a fine-grained information model named PhenoSSU (Semantic Structured Unit of Phenotypes) [bib_ref] Constructing high-fidelity phenotype knowledge graphs for infectious diseases with a fine-grained semantic..., Deng [/bib_ref] , which can accurately characterize phenotype information from medical guidelines with 12 attributes from SNOMED CT (Systematized Nomenclature of Medicine-Clinical Terms). To explore an effective strategy for deep-phenotyping Chinese EHRs, we tried to annotate some Chinese EHRs with the PhenoSSU model. During the annotation process, some linguistic patterns of PhenoSSU instances were found to frequently occur in the free text of Chinese EHRs. For example, there is a linguistic pattern of "attribute + attribute + attribute + phenotype" in a given Chinese sentence "患者反复 出现(attribute)剧烈(attribute)腹部(attribute)疼痛(phenotype)" (English translation: "patients with repeated severe abdominal pain"). If the linguistic patterns of PhenoSSU instances could be effectively learned from the corpus of Chinese EHRs, it would be possible to perform deep phenotyping of Chinese EHRs by scanning linguistic patterns of PhenoSSU instances. Therefore, how to effectively learn linguistic patterns of PhenoSSU instances from the corpus of Chinese EHRs has become an important question.
Although linguistic patterns of PhenoSSU instances can be observed and summarized manually, this is a time-consuming process that depends on experienced experts. In the field of linguistic pattern mining, the Apriori-based method is one of the most representative algorithms, which was based on the principle of frequency counts of keyword occurrences [bib_ref] Review of Apriori based frequent itemset mining solutions on big data, Fard [/bib_ref]. The Apriori algorithm is well suited to simple linguistic pattern mining based on word co-occurrence. For example, a recent study used the Apriori algorithm to learn linguistic patterns of cyberbullying behaviors in a social networking service [bib_ref] Association analysis of cyberbullying on social media using Apriori algorithm, Zainol [/bib_ref]. When two keywords co-occur frequently, they are considered to constitute a potential linguistic pattern, such as the co-occurrence of "foolish" and "abuse." However, the linguistic patterns of the PhenoSSU instances are more complicated. Thus, Apriori-based methods are not competent at mining linguistic patterns of PhenoSSU instances because they cannot handle the co-occurrence of a phenotype and several attribute values simultaneously. Inspired by the work of Ofer et al, which considered biological sequences, such as DNA sequences, as human language and used advanced NLP tools to tackle biological tasks, we aimed to model Chinese EHRs as DNA-like sequences and mine linguistic patterns with advanced bioinformatics tools. In a recent review, Castellana et al [bib_ref] A comparative benchmark of classic DNA motif discovery tools on synthetic data, Castellana [/bib_ref] surveyed 16 classic DNA motif discovery tools and evaluated their ability to discover sequence motifs nested in 29 simulated sequence data sets. The MEME (Multiple Expectation Maximums for Motif Elicitation) motif discovery tool performed best among the 16 classic DNA motif discovery tools. In this study, we characterized phenotypes as "P" and attributes as "A" to transform the free text into a single-letter sequence that could be analyzed with the MEME motif discovery tool. The sequence motifs discovered in this single-letter sequence could be viewed as linguistic patterns of PhenoSSU instances in Chinese EHRs. Based on the linguistic patterns discovered in EHRs, we could identify PhenoSSU instances by recognizing linguistic patterns in free text. To summarize, the task of deep phenotyping of Chinese EHRs could be converted into two consecutive steps of sequence motif discovery and linguistic pattern recognition.
Following this idea, we aimed to identify linguistic patterns of PhenoSSU instances in Chinese EHRs with a biological sequence motif discovery tool and develop a deep-phenotyping algorithm for Chinese EHRs by scanning linguistic patterns in free text. The rest of this paper is organized as follows. The first section introduces the composition of the PhenoSSU model and its common linguistic patterns in free text. The second section introduces the method for using a biological sequence motif discovery tool to learn linguistic patterns from the corpus of Chinese EHRs. The third section introduces the method for recognizing PhenoSSU instances from Chinese EHRs based on linguistic patterns. The final section introduces a case study to illustrate the potential application of the deep-phenotyping algorithm. Although the deep-phenotyping algorithm developed in this study can only deal with Chinese EHRs, the underlying methodology can also be illuminating for other non-English-speaking countries.
# Methods
## Overview
In this study, a data-driven approach was proposed for learning linguistic patterns from Chinese EHRs. By using a pipeline of encoding the training set as a single-letter sequence and analyzing the sequence with the MEME motif discovery tool, we learned of six regular expressions and then introduced them into our pattern recognition-based algorithm for attribute prediction. The whole pipeline for the linguistic pattern-learning method is shown in [fig_ref] Figure 1: The pipeline for the linguistic pattern-learning method [/fig_ref].
## The design of the phenossu model for representing phenotype information in chinese ehrs
PhenoSSU is essentially an entity-attribute-value model consisting of phenotype terms along with standardized attributes from SNOMED CT. Compared with two commonly used information models named CEM and FHIR, the PhenoSSU model is more suitable for the task of deep phenotyping for two reasons. First, it has been shown that the PhenoSSU model is better at representing phenotype information in medical text than CEM and FHIR models [bib_ref] Constructing high-fidelity phenotype knowledge graphs for infectious diseases with a fine-grained semantic..., Deng [/bib_ref]. Second, the PhenoSSU model puts more focus on characterizing phenotype traits with standardized attribute and value sets; as well, the attribute and value sets of the PhenoSSU model are easier to adjust according to the study-specific corpus.
To develop a fine-grained annotated corpus, 1000 Chinese EHRs of respiratory system diseases were manually annotated based on the PhenoSSU model, whose design was based on infectious diseases with a large proportion of respiratory diseases [bib_ref] Constructing high-fidelity phenotype knowledge graphs for infectious diseases with a fine-grained semantic..., Deng [/bib_ref]. These 1000 Chinese EHRs were obtained from the EHR database of the Iiyi website; all of the patients' private information in these EHRs have been masked by the Iiyi website.
During manual annotation, we optimized the attributes included in the PhenoSSU model to make them suitable for Chinese EHRs. The optimized PhenoSSU model contained 10 attributes, which could be further divided into two subtypes: (1) attributes for phrase-based phenotypes, such as "heavy cough" or "fever," including assertion, severity, temporal pattern, laterality, spatial pattern, quadrant pattern, and body location, and (2) attributes for logic-based phenotypes, such as "WBC [white blood cell] 12.5 × 10 9 /L," including specimen, analyte, and abnormality. The composition of the PhenoSSU model is shown in [fig_ref] Figure 1: The pipeline for the linguistic pattern-learning method [/fig_ref] and [fig_ref] Table 1: Six regular expressions based on linguistic patterns of the Chinese electronic health... [/fig_ref] in Multimedia Appendix 1, as well as the definitions, typical values, and SNOMED CT codes of attributes included in the model.
The phenotype information in free text could be structurally represented by the PhenoSSU model. For example, the description "a sudden severe pain in the right-lower abdomen" could be represented as a PhenoSSU instance consisting of the phenotype concept "pain," the assertion attribute "present," the temporal pattern attribute "acute," the severity attribute "severe," the quadrant pattern attribute "right-lower," and the body location attribute "abdomen." Meanwhile, logic-based phenotypes (ie, qualitative and quantitative test results) were also included in the PhenoSSU model. For example, "WBC 12.5 × 10 9 /L" could be represented as a PhenoSSU instance consisting of the analyte "WBC" and the abnormality attribute "abnormality: higher," which was combined and normalized as a concept of the "increased blood leukocyte number (414478003)" in SNOMED CT [fig_ref] Figure 2: Free-text phenotype descriptions and linguistic patterns [/fig_ref] , A). The relevant knowledge came from our previous study, LATTE (transforming lab test results) [bib_ref] LATTE: A knowledge-based method to normalize various expressions of laboratory test results..., Jiang [/bib_ref] , which was integrated into this work, including sample sources, analyte names, and reference ranges for 1098 laboratory tests. Detailed information about the knowledge base is shown in [fig_ref] Figure 2: Free-text phenotype descriptions and linguistic patterns [/fig_ref] in Multimedia Appendix 1.
Based on the annotation guideline of the PhenoSSU model in our previous work, two Chinese authors with medical backgrounds (LC and SL) manually annotated these medical records independently. Annotations were made on the brat rapid annotation tool platform [bib_ref] brat: A web-based tool for NLP-assisted text annotation, Stenetorp [/bib_ref]. The initial annotating agreement measured with the Cohen κ statistic was 0.851. All inconsistent annotations were decided by the project supervisor (TJ).
During annotation, we found some linguistic patterns of PhenoSSU instances in the EHR text. For example, the description of a phrase-based phenotype, "右下腹部突发剧烈 疼痛" (English translation: "a sudden severe pain in the right-lower abdomen"), could be summarized as "attribute (right-lower) + attribute (abdomen) + attribute (acute) + attribute (severe) + phenotype (pain)." Similarly, the description of logic-based phenotypes had common linguistic patterns in free text, such as "analyte (WBC) + number (12.5 × 10 9 ) + unit (cells/ L)" [fig_ref] Figure 2: Free-text phenotype descriptions and linguistic patterns [/fig_ref] , B). If we can mine linguistic patterns of PhenoSSU instances from Chinese EHRs, it would be possible to develop pattern recognition-based deep phenotyping.
## Learning linguistic patterns of phenossu instances from chinese ehrs using meme: workflow
## Overview
In order to learn linguistic patterns of PhenoSSU instances from the Chinese EHR corpus, 1000 annotated Chinese EHRs in the study were divided into a training set (n=700, 70%) and test set (n=300, 30%). The workflow of linguistic pattern mining is shown in [fig_ref] Figure 3: The workflow of learning linguistic patterns of the PhenoSSU model from the... [/fig_ref] , which includes two stages: pattern discovery and pattern enrichment. In the stage of linguistic pattern discovery, we used the MEME motif discovery tool, which solves the problem of motif mining with a maximum likelihood method [bib_ref] The MEME suite, Bailey [/bib_ref] to obtain seed linguistic patterns of PhenoSSU instances. In the stage of linguistic pattern enrichment, a semiautomatic method was developed to check and fill linguistic pattern gaps. Through pattern discovery and enrichment, we built a linguistic pattern library of PhenoSSU instances.
## Stage 1: linguistic pattern discovery
First, free text in the training set was encoded into single-letter sequences. To represent EHRs as the input of the MEME motif discovery tool, we encoded them as single-letter sequences with the following criteria: the phenotype (ie, "fever" and "cough") was encoded as "P" and the attribute (ie, "severe") was encoded as "A." In the description of phrase-based phenotypes, "P" and "A" could be directly recognized in the original text. However, to calculate the abnormality of a logic-based phenotype, we need to combine the specimen ("S"), analyte ("L"), number ("N"), and unit ("U"). Specifically, the source of laboratory examination (ie, "blood" and "urine") was encoded as "S," the analyte (ie, "leukocyte") was encoded as "L," the number was encoded as "N" (ie, "37"), and the unit (ie, "°C") was encoded as "U." Meanwhile, the punctuation (ie, a comma) was encoded as "C," and other information was encoded as "O." In this study, EHRs were encoded using the FlashText tool, a tool for string-based concept recognition and replacement. FlashText can find and replace keywords based on the trie dictionary data structure, which is 82 times faster than regular expressions. Because of its efficiency in processing text, we chose the FlashText tool for encoding text as single-letter sequences. Note that FlashText can retain the index of the strings in the original text. For example, the free-text description "患 者主诉 Second, the MEME motif discovery tool was used to mine motifs in the single-letter sequence. The pipeline of MEME motif discovery is composed of three steps: finding starting points, maximizing the likelihood expectation, and scoring the discovered motifs.
The input was a set of unaligned sequences, and the output was a list of probable motifs. The statistical significance of the motifs in MEME was evaluated by the E value, which is based on the log-likelihood ratio. The settings of the MEME motif discovery tool were optimized as follows:
1. Motif discovery mode: classic mode. In classic mode, only one sequence needs to be provided. The algorithm will find the repeated sequence fragments in the sequence by likelihood ranking. 2. Select the site distribution: any number of repetitions. This option means selecting motifs that occur repeatedly. 3. How wide can motifs be: from 2 to 30. This number is the width (ie, characters in the sequence pattern) of a single motif. MEME can choose an optimal width of each motif individually by using a heuristic function. In the process, there were some motifs containing "O" (ie, other information), which was irrelevant to phenotype descriptions. Therefore, we separated out the motifs with the letter "O" to generate sequence segments that may represent linguistic patterns of PhenoSSU instances.
Third, we built regular expressions based on the discovered motifs. To make the motifs available in our algorithm, regular expressions were built. For example, we built a regular expression "A+P" based on motifs or sequence segments generated from motifs like "AP," "AAP," "AAAP," and "AAAAP."
## Stage 2: linguistic pattern enrichment
In this stage, a linguistic pattern recognition-based method was first used to automatically recognize PhenoSSU instances from Chinese EHRs in the training set. The workflow of linguistic pattern recognition is shown in [fig_ref] Figure 4: The workflow of recognizing PhenoSSU instances from free text via linguistic pattern... [/fig_ref] , which includes the following steps:
1. Encoding text as single-letter sequences. For example, the description "右下腹部突发剧烈疼痛" (English translation: "a sudden severe pain in the right-lower abdomen") was encoded as the single-letter sequence "AAAAP." The FlashText tool could record the position index of Chinese characters in every single letter, making it possible to map single letters to the original text. An example of the position index recording is shown in [fig_ref] Figure 4: The workflow of recognizing PhenoSSU instances from free text via linguistic pattern... [/fig_ref] in Multimedia Appendix 1. 2. Scanning the single-letter sequence with the linguistic patterns. In this case, "AAAAP" matched the linguistic pattern "A + P" perfectly, meaning that the four attributes were associated with the phenotype. 3. Mapping these letters to the original text by index. A: right-lower; A: abdomen; A: acute; A: severe; P: pain. 4. Filling phenotypes and associated attributes in the PhenoSSU model. Finally, the description "右下腹部突发 剧烈疼痛" could be transformed into a PhenoSSU instance consisting of the phenotype "pain," the assertion attribute "present," the temporal pattern attribute "acute," the severity attribute "severe," the quadrant pattern attribute "right-lower," and the body location attribute "abdomen."
Based on the above steps, we discovered the unrecognized PhenoSSU instances by comparing the automatically recognized instances with manual annotation. For example, the description "没有出现(A)发热(P),(C)乏力(P)" (English translation: "No fever or fatigue") could be encoded as "APCP," in which "AP" matched the regular expression ("A + P") in our pattern library. By mapping to the original text, "没有出现发热，乏力" was transformed into a PhenoSSU instance consisting of the phenotype "fever" and the assertion attribute "absent." However, "absent" was also the attribute of the phenotypes "diarrhea" and "weight loss," which were not recognized by the algorithm.
Finally, to check why these PhenoSSU instances were not recognized, all of them were encoded as single-letter sequences, which could be scanned with linguistic patterns. If no pattern matched, we collected such sequences to build new regular expressions and add them to the linguistic pattern library. In this example, sequences such as "APCPCP" were enriched into a regular expression "(A + P (CP) +)."
## Recognizing phenossu instances from chinese ehrs: workflow
The recognition of PhenoSSU instances could be divided into two subtasks: entity recognition and attribute prediction. To find the best strategy for the two tasks, it was essential to compare our proposed method with current state-of-the-art methods.
The first subtask was entity recognition, which aimed to recognize the text spans corresponding to phenotype and attribute entities. For the subtask of named entity recognition from Chinese EHRs, the Bidirectional Encoder Representations from Transformers (BERT)-Bi-LSTM-CRF model has proven its effectiveness in the CCKS (China Conference on Knowledge Graph and Semantic Computing) 2018 Task 1: Named Entity Recognition in Chinese Electronic Medical Records, which achieved the best F1 score of 91.43 [bib_ref] Overview of CCKS 2018 Task 1: Named Entity Recognition in Chinese Electronic..., Zhang [/bib_ref]. Therefore, we compared algorithm performances of the BERT-Bi-LSTM-CRF model and the classic dictionary-based method in this study. The parameters of the BERT model were trained with the Kashgari package in Python (version 3.6.1; Python Software Foundation). In the dictionary-based method, the knowledge base of phenotypes was derived from the Chinese translations of the International Classification of Diseases, 10th Revision and 11th Revision, and the Human Phenotype Ontology (details in in Multimedia Appendix 1). Further, the knowledge base of attribute trigger words was from the annotation of the training set. Entity recognition, combined with the other coding rules, was applied to encode free text as single-letter sequences, which would be used in the subsequent attribute prediction subtask.
The phenotype's attribute recognition was the second subtask, which aimed to predict appropriate values for the [bib_ref] Semantic integration of clinical laboratory tests from electronic health records for deep..., Zhang [/bib_ref]
## Evaluation of algorithm performance for recognizing phenossu instances
To evaluate the algorithm's performance for recognizing PhenoSSU instances, we used the evaluation metrics in SemEval (Semantic Evaluation) 2015 Task 14: Analysis of Clinical Text [bib_ref] SemEval-2015 Task 14: Analysis of Clinical Text, Elhadad [/bib_ref].
In the subtask of entity recognition, the F1 score was taken as the evaluation metric. When a predicted entity word entirely coincided with a gold-standard text span, it was considered as a true positive. The precision metric was calculated as the fraction of correctly predicted entities among all entities identified by the algorithm, and the recall metric was calculated as the fraction of correctly predicted entities among all entities identified by the annotators. The F1 score was calculated as the harmonic mean of precision and recall.
In the subtask of attribute prediction, the average accuracy and weighted average accuracy were taken as the evaluation metrics because the weighted average accuracy thoroughly considered the distribution of each attribute value in the corpus, which could better evaluate those attribute values with little distribution.
For the evaluation at the PhenoSSU-instance level, we used the combination of the F1 score for entity recognition and weighted average for attribute prediction. A PhenoSSU instance was considered correct when the phenotype and associated attribute values annotated by the algorithm were the same as the corresponding PhenoSSU instance annotated by experts.
# Ethical considerations
The 1000 Chinese EHRs of respiratory system diseases used in this study were obtained from the EHR database of the Iiyi website. No ethics approval was needed because the data from downloaded EHRs, including patients' private information, were all masked by the Iiyi website.
# Results
## Linguistic patterns of phenossu instances learned from chinese ehrs
A total of 51 sequence motifs were discovered from the Chinese EHRs in the training set (details are shown in [fig_ref] Figure 5: Determining the best strategy for recognizing PhenoSSU instances [/fig_ref] in Multimedia Appendix 1). Based on the 51 motifs, we built six regular expressions [fig_ref] Table 1: Six regular expressions based on linguistic patterns of the Chinese electronic health... [/fig_ref] , namely linguistic patterns of the PhenoSSU instances in the Chinese EHRs. Among the regular expressions of phrase-based phenotypes, "AP +" appeared most frequently. The most common description of this regular expression was "absent" plus phenotypes, which could be used for differential diagnosis in clinical practice. The second frequent regular expression was "A + P," which usually corresponded to a detailed description of phenotypes, such as "body location + severity + phenotype." There were also complex linguistic patterns to be generalized as "A × PC × A +," for example, "严重(A)咳嗽(P)，(C)呈持续性(A)" (ie, severe cough, consistently). Among the regular expressions of logic-based phenotypes, the most typical was "S × LNU," such as the description "WBC 12 × 10 9 /L." There were also linguistic patterns that directly interpreted laboratory examination results: "S × LR [results of laboratory examination]," such as "血糖升 高" (ie, high blood glucose). The above results suggest that there are inherent linguistic patterns in Chinese EHRs. The detailed frequency of linguistic patterns is shown in in Multimedia Appendix 1.
In this study, six regular expressions were learned from 700 Chinese EHRs in the training set. However, the size of the training set could be smaller than 700 in order to build the six regular expressions. To explore the potential smallest size of the training set, we conducted an experiment to explore the minimum number of EHRs that could match all six regular expressions. In the experiment, we randomly selected EHRs from the training set with stepwise increased data size, which were scanned with the six regular expressions. When all six regular expressions could be matched, that data size was recorded. This process was repeated 1000 times to calculate the mean and SD of the EHR sums that covered the six regular expressions. Results showed that in a mean of 134 (SD 9.7) EHRs, the six regular expressions could be matched. We did not use the pattern discovery method illustrated in this study because there was a semiautomatic step in the method.
Repeating the pattern discovery method 1000 times would be time-consuming. A line graph was plotted to show five examples among all 1000 tests [fig_ref] Figure 6: The comparison of PhenoSSU instances extracted from the clinical guidelines and electronic... [/fig_ref] in Multimedia Appendix 1).
## Example in chinese (english translation) phenotype category and regular expressions
Phrase-based phenotypes "无/A咳嗽/P、/C发热/P" (no cough or fever) re.compile a ("A b +P c (C d P)+") "严重/A腹痛/P腹泻/P" (severe abdominal pain and diarrhea) re.compile("AP+") "右下腹/A严重/A疼痛/P" (severe right-lower abdominal pain) re.compile("A+P") "咳嗽/P，/C呈持续性/A" (cough, consistently) re.compile("A×PC×A+")
## Logic-based phenotypes
"白细胞/L 12×10 9 /N /L/U" (WBC i 12 × 10 9 /L) re.compile("S e ×L f N g U h ") "血/S糖/L升高/R" (high blood glucose) re.compile("S×LR j " ) a re.compile: a Python method used to compile a regular expression pattern.
## The best strategy for recognizing phenossu instances
Based on the linguistic patterns of Chinese EHRs, we developed a pattern recognition-based method to identify PhenoSSU instances. To find the best strategy for recognizing PhenoSSU instances, we developed and compared different methods in the subtasks of entity recognition and attribute prediction. The results in [fig_ref] Figure 5: Determining the best strategy for recognizing PhenoSSU instances [/fig_ref] show that the best strategy was to recognize entities using the deep learning-based method and then predict the attribute values using the pattern recognition-based method.
Specifically, in the entity recognition subtask, the method of deep learning (ie, BERT-Bi-LTSM-CRF) achieved the best performance, with an F1 score of 0.898 [fig_ref] Figure 5: Determining the best strategy for recognizing PhenoSSU instances [/fig_ref]. As a comparison, the dictionary-based method achieved an F1 score of 0.804. In the subtask of attribute prediction, the pattern recognition-based method had the best performance, with an accuracy of 0.977 and a weighted average of 0.940 [fig_ref] Figure 5: Determining the best strategy for recognizing PhenoSSU instances [/fig_ref] , C). The SVM-based method achieved an accuracy of 0.783 and a weighted average of 0.709. The deep-phenotyping algorithm for Chinese EHRs had an overall accuracy of 0.844 on the test set. The detailed performances of the two models for predicting attribute values are shown in in Multimedia Appendix 1.
## Case study: exploring the real-world evidence that deep-phenotyping ehrs can update knowledge in guidelines
With the pattern recognition algorithm, we could effectively structure phenotype information in Chinese EHRs. To demonstrate the potential application of deep phenotyping, a case study was conducted to update clinical guidelines by information retrieval of EHRs. In the case study, we selected the latest Chinese clinical guideline and 300 Chinese EHRs of chronic bronchitis. To recognize PhenoSSU instances from the guideline and the EHRs, we used the optimized hybrid strategy mentioned previously.
A total of 9 and 29 PhenoSSU instances were identified from the clinical guideline and the EHRs of chronic bronchitis, respectively (details are shown in Tables S5-S7 in Multimedia Appendix 1). The 9 PhenoSSU instances identified in the clinical guideline appeared in the EHRs, which meant another 20 PhenoSSU instances in the EHRs were not covered in the clinical guideline. For example, "cough: chronic" and "cough: recurrent" both appeared in the clinical guideline and the EHRs. However, the current guideline could not give suggestions to accurately diagnose patients with occasional cough or severe cough as having chronic bronchitis [fig_ref] Figure 6: The comparison of PhenoSSU instances extracted from the clinical guidelines and electronic... [/fig_ref]. This real-world evidence hints at the feasibility of updating knowledge in clinical guidelines through deep phenotyping of large-scale EHRs.
# Discussion
## Principal findings
In this study, we developed a simple but effective strategy to perform deep phenotyping of Chinese EHRs. The core of this strategy is learning linguistic patterns of PhenoSSU instances with a motif discovery tool from the field of bioinformatics. According to this research, biological sequence motif discovery tools could be used to effectively identify linguistic patterns of phenotype descriptions from medical texts after encoding them as DNA-like sequences. Meanwhile, the process of identifying linguistic patterns does not require too much annotation data; thus, our strategy is suitable for low-resource scenarios of deep-phenotyping Chinese EHRs.
This study was a preliminary attempt to use bioinformatics tools to tackle problems in medical informatics. By modeling natural language as single-letter sequences, it is possible that other advanced tools for analyzing biological sequences could also be used for processing natural language. For example, some researchers in the NLP field have applied a classic informatics algorithm, named the Basic Local Alignment Search Tool (BLAST), [bib_ref] NCBI BLAST: A better web interface, Johnson [/bib_ref] to the text reuse detection task [bib_ref] A BLAST-based, language-agnostic text reuse algorithm with a MARKUS implementation and sequence..., Vierthaler [/bib_ref]. In Vesanto's work, the 23 most-used English letters in the data set were calculated to form a simple one-to-one mapping between English letters and arbitrary amino acids. In this way, text was encoded into single-letter sequences that BLAST could handle to calculate similarities between texts. It is believed that future communications between bioinformatics and medical informatics will become more frequent [bib_ref] A scoping review of 'big data', 'informatics', and 'bioinformatics' in the animal..., Ouyang [/bib_ref].
It can be concluded from this study that there exist some regular linguistic patterns for phenotype narratives in Chinese EHRs.
The origin of these linguistic patterns may be the common writing habits of clinicians who try to save time by recording clinical information faithfully in as few words as possible. The reason our strategy does not require large annotation samples is that it uses the inner knowledge of linguistic patterns. As we know, data-hungry strategies, such as machine learning and deep learning, require many training samples to effectively identify patterns from data. However, there are many low-resource scenarios in practice that lack sufficient annotation samples for machine learning or deep learning. This is perhaps the reason why the majority (60%) of NLP studies in the medical domain have continued to use a knowledge-based approach rather than a machine learning-based approach [bib_ref] Clinical information extraction applications: A literature review, Wang [/bib_ref]. In recent years, researchers have become increasingly focused on integrating machine learning with human knowledge, which is expected to become a new paradigm to deal with low-resource scenarios in medical informatics [bib_ref] Towards realizing the vision of precision medicine: AI based prediction of clinical..., De Jong [/bib_ref].
# Limitations
One limitation of this study was that linguistic patterns were learned from EHRs of respiratory diseases, which may not be applicable to other diseases. In addition, limited by the data size, the linguistic patterns in our study might be incomplete. In the future, we will continue to improve the algorithm with more Chinese EHRs from different hospital departments.
# Conclusions
We developed a simple but effective strategy to perform deep phenotyping of Chinese EHRs with limited fine-grained annotation data. Our work will promote the second use of Chinese EHRs and bring inspiration to other non-English-speaking countries.
[fig] Figure 1: The pipeline for the linguistic pattern-learning method. A: attribute; C: punctuation; EHR: electronic health record; MEME: Multiple Expectation Maximums for Motif Elicitation; O: other information; P: phenotype; PhenoSSU: Semantic Structured Unit of Phenotypes; re.compile: a Python method used to compile a regular expression pattern; SNOMED CT: Systematized Nomenclature of Medicine-Clinical Terms. [/fig]
[fig] Figure 2: Free-text phenotype descriptions and linguistic patterns. A. Examples of structuring free text by the PhenoSSU model. B. Examples of linguistic patterns in free text. A: attribute; L: analyte; N: number; P: pain; PhenoSSU: Semantic Structured Unit of Phenotypes; U: unit; WBC: white blood cell. [/fig]
[fig] Figure 3: The workflow of learning linguistic patterns of the PhenoSSU model from the corpus of Chinese electronic health records (EHRs). PhenoSSU: Semantic Structured Unit of Phenotypes; re.compile: a Python method used to compile a regular expression pattern. [/fig]
[fig] Figure 4: The workflow of recognizing PhenoSSU instances from free text via linguistic pattern recognition. The numbers within the square brackets represent the position indexes of single letters in the original text. A: attribute; P: phenotype; PhenoSSU: Semantic Structured Unit of Phenotypes; re.compile: a Python method used to compile a regular expression pattern. [/fig]
[fig] Figure 5: Determining the best strategy for recognizing PhenoSSU instances. A. The workflow of recognizing PhenoSSU instances from free text. B. The performance comparison between the dictionary-based method and the deep learning-based method in identifying phenotype concepts. C. The performance comparison between the SVM-based method and the pattern recognition-based method in recognizing a phenotype's attributes. PhenoSSU: Semantic Structured Unit of Phenotypes; SNOMED CT: Systematized Nomenclature of Medicine-Clinical Terms; SVM: support vector machine. [/fig]
[fig] Figure 6: The comparison of PhenoSSU instances extracted from the clinical guidelines and electronic health records (EHRs) of chronic bronchitis. PhenoSSU: Semantic Structured Unit of Phenotypes. [/fig]
[table] Table 1: Six regular expressions based on linguistic patterns of the Chinese electronic health record corpus in this study. [/table]
|
Natural products for cancer chemotherapy
For over 40 years, natural products have served us well in combating cancer. The main sources of these successful compounds are microbes and plants from the terrestrial and marine environments. The microbes serve as a major source of natural products with antitumour activity. A number of these products were first discovered as antibiotics. Another major contribution comes from plant alkaloids, taxoids and podophyllotoxins. A vast array of biological metabolites can be obtained from the marine world, which can be used for effective cancer treatment. The search for novel drugs is still a priority goal for cancer therapy, due to the rapid development of resistance to chemotherapeutic drugs. In addition, the high toxicity usually associated with some cancer chemotherapy drugs and their undesirable side-effects increase the demand for novel anti-tumour drugs active against untreatable tumours, with fewer side-effects and/or with greater therapeutic efficiency. This review points out those technologies needed to produce the anti-tumour compounds of the future.
# Introduction
Natural products have been an overwhelming success in our society. The use of plant and microbial secondary metabolites has aided in doubling of our life span in the 20th century. They have reduced pain and suffering, and revolutionized medicine by enabling the transplantation of organs. Since their chemical diversity is based on biological and geographical diversity, the entire globe is explored for bioprospecting by researchers. Researchers have had easy access to terrestrial life from which most of the pharmaceutically successful natural products originate. However, the ocean hosts a vast repertoire of life forms brimming with natural products of potential pharmaceutical importance. Marine bioprospecting is a relatively new phenomenon; thus, marine life is a relatively unexplored area of opportunity. New methods are being developed to grow the so-called 'unculturable' microbes from both the soil and the sea. Most biologically active natural products are secondary metabolites with complex structures. In some cases, the natural product itself can be used, but in others, derivatives made chemically or biologically are the molecules used in medicine. Biosynthetic pathways are often genetically manipulated to yield the desired product. With the advent of combinatorial biosynthesis, thousands of new derivatives can now be made by this biological technique, which is complementary to combinatorial chemistry.
Secondary metabolism evolved in nature in response to needs and challenges of the natural environment. Nature has been continually carrying out its own version of combinatorial chemistry [bib_ref] The combinatorial chemistry of nature, Verdine [/bib_ref] for over the three billion years in which bacteria have inhabited the earth [bib_ref] Evidence for life on earth more than 3850 million years ago, Holland [/bib_ref]. During that time, there has been an evolutionary process going on in which producers of secondary metabolites evolved according to their local environments. If the metabolites were useful to the producing species, the biosynthetic genes were retained and genetic modifications further improved the process. Combinatorial chemistry practised by nature is much more sophisticated than combinatorial chemistry in the laboratory, yielding exotic structures rich in stereochemistry, concatenated rings and reactive functional groups [bib_ref] The combinatorial chemistry of nature, Verdine [/bib_ref]. As a result, an amazing variety and number of products have been found in nature. This natural wealth is tapped for drug discovery using high-throughput screening and fermentation, mining genomes for cryptic pathways, and combinatorial biosynthesis to generate new secondary metabolites related to existing pharmacophores. The success of the pharmaceutical industry depends on the combination of complementary technologies such as natural product discovery, high-throughput screening, genomics, proteomics, metabolomics and combinatorial biosynthesis.
About a million natural products are known. The number produced by plants has been estimated to be between 500 000 and 600 000 [bib_ref] The value of undiscovered pharmaceuticals in tropical forests, Mendelson [/bib_ref]. The structures of 160 000 natural products were already elucidated by the late 1990s, a value growing by 10 000 per year [bib_ref] Statistical investigation into structural complementarity of natural products and synthetic compounds, Henkel [/bib_ref]. Of these, about 100 000 are from plants. There are more than 20 000 microbial secondary metabolites [bib_ref] Diversifying microbial natural products for drug discovery, Knight [/bib_ref]. With regard to biological activity, there are about 200 000 to 250 000 biologically active products (active and/or toxic). About 100 000 secondary metabolites of molecular weight less than 2500 have been characterized. About half are produced by microbes and the other half by plants [bib_ref] Marine microorganisms: a new biomedical resource, Fenical [/bib_ref].
Many of the terrestrial natural products succeeded in clinical trials and have effectively served medicine, some for over 50 years. Over 60% of approved and pre-NDA candidates are either natural products or related to them, not including biologicals such as vaccines and monoclonal antibodies [bib_ref] Natural products in drug discovery and development, Cragg [/bib_ref] [bib_ref] Natural products as sources of new drugs over the period 1981-2002, Newman [/bib_ref]. Six per cent are natural products, 27% are derivatives of natural products, 5% are synthetic with natural product pharmacophores, and 23% are synthetic mimics of natural products. Almost half of the best selling pharmaceuticals are natural or are related to natural products. In early 2008, there were 225 natural product-based drugs in various testing procedures such as preclinical, clinical phases I to III, and preregistration [bib_ref] Natural products in drug discovery, Harvey [/bib_ref]. Of these, 108 were from plants, 61 were semisynthetic, 25 were from bacteria, 24 from animals and 7 were fungal in origin. Of the 18 000 known marine natural products, 22 of these or their chemical derivatives are in clinical trials.
With regard to the structures of natural products, the estimated number of known isoprenoids (including terpenoids and carotenoids) is 50 000 [bib_ref] Prospects for the bioengineering of isoprenoid biosynthesis, Mccaskill [/bib_ref]. Terpenes number 30 500 and as pharmaceuticals had a market of $12 billion in 2002 [bib_ref] Pharmacognosy in the new millennium: leadfinding and biotechnology, Verpoorte [/bib_ref] [bib_ref] Plants and human health in the twenty-first century, Raskin [/bib_ref]. Other major categories include polyketides, including macrolides, non-ribosomal peptides, etc. About 4000 of the known natural products are halogenated.
Cancer is a major group of diseases. There were 1.5 million new cases in the USA in 2008. The global market for anti-tumour agents was $60 billion in 2007. Natural products are the most important anti-cancer agents. Three quarters of anti-tumour compounds used in medicine are natural products or related to them. Of the 140 anti-cancer agents approved since 1940 and available for use, over 60% can be traced to a natural product. Of the 126 small molecules among them, 67% are natural in origin [bib_ref] Plants as a source of anti-cancer agents, Cragg [/bib_ref]. In 2000, 57% of all drugs in clinical trials for cancer were either natural products or their derivatives [bib_ref] Antineoplastic agents from natural sources: achievements and future directions, Cragg [/bib_ref]. From 1981 to 2002, natural products were the basis of 74% of all new chemical entities for cancer. Of the 225 natural product-based drugs in various stages of clinical testing in 2008 mentioned above [bib_ref] Natural products in drug discovery, Harvey [/bib_ref] , the therapeutic categories targeted included 86 for cancer.
Compounds with anti-tumour activity belong to several structural classes such as anthracyclines, enediynes, indolocarbazoles, isoprenoids, polyketide macrolides, non-ribosomal peptides including glycopeptides, and others. Most of the polyketides are produced by bacteria and fungi [bib_ref] Modular enzymes, Rawls [/bib_ref] [bib_ref] Electronic structure near the Fermi surface in the quasi-one-dimensional conductor Li0.9Mo6O17, Xue [/bib_ref]. They include a number of anti-tumour drugs such as taxol, which is made by both plants and fungi. Halogenated anti-tumour candidates include salinosporamide A and rebeccamycin [bib_ref] Halogenation strategies in natural product biosynthesis, Neumann [/bib_ref].
The anti-tumour compounds act by several mechanisms such as inducing apoptosis (programmed cell death) through DNA cleavage mediated by topoisomerase I or II inhibition, mitochondrial permeabilization, inhibition of key enzymes involved in signal transduction (e.g. proteases), or cellular metabolism, and by inhibiting tumour-induced angiogenesis (recruitment of new blood vessels).
## Microbial anti-tumour products
A great number of anti-tumour compounds are natural products, mainly produced by microorganisms, or their derivatives. In particular, actinomycetes are the producers of a large number of natural products with anti-tumour properties, many of which also have antimicrobial activity. Thus, a broad screening of antibiotically active molecules for antagonistic activity against organisms other than microorganisms was proposed in the 1980s in order to yield new and useful lives for 'failed antibiotics'. This resulted in the development of a large number of simple in vitro laboratory tests, e.g. enzyme inhibition screens [bib_ref] Low-molecular-weight enzyme inhibitors of microbial origin, Umezawa [/bib_ref] to detect, isolate and purify useful compounds. As a result, we entered into a new era in which microbial metabolites were applied to diseases heretofore only treated with synthetic compounds, and much success was achieved. One such area was that of anti-tumour agents.
In their review on the use of microbes to prescreen potential anti-tumour compounds, [bib_ref] Microbial prescreens for anticancer activity, Newman [/bib_ref] concluded that microorganisms have played an important role in identifying compounds with therapeutic benefit against cancer. Most of the important compounds used for chemotherapy of tumours are microbially produced antibiotics or their derivatives. Some of the microbial anti-tumour agents are shown in [fig_ref] Table 1: Some microbial anti-tumour compounds [/fig_ref]. One of the earliest applications of a microbial product was actinomycin D for Wilm's tumour in children. Use of this compound against stage I or stage II Wilm's tumour has resulted in a 90% survival rate [bib_ref] H. Boyd Woodruff (b. 1917): antibiotics hunter and distinguished soil microbiologist, Chung [/bib_ref]. The structures of some of the most useful anti-tumour com-pounds are shown in [bib_ref] Genetic manipulation of antitumor-agent biosynthesis to produce novel drugs, Salas [/bib_ref]. Also used for anti-tumour therapy is the enzyme L-asparaginase.
Bleomycin is a glycopeptide produced by Streptoalloteichus hindustanus. It is used for squamous cell carcinomas, Hodgkin's lymphomas and testis tumours. A derivative of the bleomycin family, pingyangmycin, has been used in cancer therapy in China since 1978 [bib_ref] Antitumor antibiotic pingyangmycin: research and clinical use for 40 years, Zhen [/bib_ref]. Another bleomycin derivative, Blenoxane, is used clinically with other compounds against lymphomas, skin carcinomas and tumours of the head, neck and testicles [bib_ref] Functional characterization of tlmH in Streptoalloteichus hindustanus E465-94 ATCC 31158 unveiling new..., Tao [/bib_ref].
Derivatives (analogues) are usually made chemically or by manipulation of fermentation conditions. For example, addition of KBr to the rebeccamycin producer, Saccharothrix aerocolonigenes, yielded a brominated rebeccamycin [bib_ref] Isolation of a bromo analog of rebeccamycin from Saccharothrix aerocolonigenes, Lam [/bib_ref]. Addition of DL-fluorotryptopan yielded two new fluorinated rebeccamycins and addition of DL-fluorotryptophan led to a third [bib_ref] Production, isolation and structure determination of novel fluoroindolocarbazoles from Saccharothrix aerocolonigenes ATCC..., Lam [/bib_ref]. In clinical testing are the rebeccamycin derivative edotecarin and the geldanomycin derivative 17-allylaminogeldanomycin. Interestingly, a fourthgeneration tetracycline known as SF 2575, produced by Streptomyces sp., has low antibiotic activity but a high level of activity against P388 leukaemia cells in vitro and many other types of cancer cells [bib_ref] Biochemical analysis of the biosynthetic pathway of an anticancer tetracycline SF2575, Pickens [/bib_ref]. It appears to act against DNA topoisomerases I and II, the target of camptothecins and doxorubicin respectively.
A dramatic example of the power of natural products in the cure of cancer is that of metastatic testicular cancer. Although this type of cancer is rather uncommon, i.e. it was responsible for only 1% of male malignancies in the USA, it did cause 80 000 cases in the year 2000. Indeed, it is the most common carcinoma in men aged 15-35. The cure rate for disseminated testicular cancer was 5% in 1974; today it is 90% mainly due to the use of a triple combination of the microbial product bleomycin, the plant compound etoposide and the synthetic agent cisplatin [bib_ref] Cisdiamminedichloroplatinum, vinblastine, and bleomycin combination chemotherapy in disseminated testicular cancer, Einhorn [/bib_ref].
## Anthracyclines
Anthracyclines are among the most well-known clinically used anti-tumour agents, especially daunorubicin (daunomycin), doxorubicin (14-hydroxydaunorubicin); adriamycin, carminemycin and aclarubicin. The biosynthesis of daunorubicin and doxorubicin are depicted in 2 [bib_ref] Biochemistry, molecular biology and protein-protein interactions in daunorubicin/ doxorubicin biosynthesis, Strohl [/bib_ref].
A novel anthracycline, 11-hydroxyaclacinomycin A, was produced by cloning the doxorubicin resistance gene and the aklavinone 11-hydroxylase gene dnrF from the doxorubicin producer, Streptomyces peucetius ssp. caesius, into the aclacinomycin A producer [bib_ref] Expression of Streptomyces peucetius genes for doxorubicin resistance and aklavinone 11-hydroxylase in..., Hwang [/bib_ref]. The hybrid molecule showed greater activity against leukaemia and melanoma than aclacinomycin A. Another hybrid molecule produced was 2″-amino-11hydroxyaclacinomycin Y, which is highly active against tumours [bib_ref] New anthracycline metabolites produced by the aklavinone 11-hydroxylase gene in Streptomyces galilaeus..., Kim [/bib_ref]. Additional anthracyclines have been made by introducing DNA from Streptomyces purpurascens into Streptomyces galilaeus, both of which normally produce known anthracyclines [bib_ref] Nucleotide sequences and expression of genes from Streptomyces purpurascens that cause the..., Niemi [/bib_ref]. Other novel anthracyclines were produced by cloning DNA from the nogalomycin producer, Streptomyces nogalater, into Streptomyces lividans and into an aclacinomycin-negative mutant of S. galilaeus [bib_ref] Production of hybrid anthracycline antibiotics by heterologous expression of Streptomyces nogalater nogalamycin..., Ylihonko [/bib_ref]. Cloning of the actI, actIV and actVII genes from Streptomyces coelicolor into the 2-hydroxyaklavinone producer, S. galilaeus 31671, yielded the novel hybrid metabolites, desoxyerythrolaccin and 1-O-methyl-desoxyerythrolaccin [bib_ref] Expression of polyketide biosynthesis and regulatory genes in heterologous streptomycetes, Strohl [/bib_ref]. Similar studies yielded the novel metabolite aloesaponarin II [bib_ref] Biosynthesis of anthraquinones by interspecies cloning of actinorhodin genes in streptomycetes: clarification..., Bartel [/bib_ref]. Epirubicin (4′-epidoxorubicin) is a semisynthetic anthracycline with less cardiotoxicity than doxorubicin [bib_ref] Synthesis and antitumor properties of new glycosides of daunomycinone and adriamycinone, Arcamone [/bib_ref]. Genetic engineering of a blocked S. peucetius strain provided a new method to produce it [bib_ref] Production of the antitumor drug epirubicin (4′-epidoxorubicin) and its precursor by a..., Madduri [/bib_ref]. The gene introduced was avrE of the avermectin-producing Streptomyces avermitilis or the eryBIV genes of the erythromycin producer, Saccharopolyspora erythrea. These genes and the blocked gene in the recipient are involved in deoxysugar biosynthesis.
## Enediynes
Enediynes are among the strongest naturally produced anti-tumour compounds but are extremely toxic due to their action of causing apoptosis in normal cells as well as in tumour cells. They include calicheamicin, dynemicin A, esparamicin, kerdarcidin and neocarzinostatin. In recent years, scientists are trying to design nontoxic enediynebased anti-tumour drugs [bib_ref] Computer design of anticancer drugs. A new enediyne warhead, Kraka [/bib_ref]. Progress towards this goal is proceeding by merging amidines with the natural enediyne dynemicin A [bib_ref] Design of a new warhead for the natural enediyne dynemicin A. An..., Kraka [/bib_ref].
## Epothilones
An unusual source of secondary metabolites are the myxobacteria, relatively large Gram-negative rods which move by gliding or creeping. They form fruiting bodies and have a very diverse morphology. Over 400 compounds had been isolated from these organisms by 2005, but the first in clinical trials were the epothilones, potential antitumour agents, which act like taxol (see Plant anti-tumour agents below) but are active versus taxol-resistant tumours [bib_ref] The microtubule-stabilizing agent discodermolide competitively inhibits the binding of paclitaxel (Taxol) to..., Kowalski [/bib_ref]. Epothilones are 16-member ring polyketide macrolide lactones produced by the myxobacterium Sorangium cellulosum, which were originally developed as antifungal agents against rust fungi [bib_ref] Pharmacognosy in the new millennium: leadfinding and biotechnology, Verpoorte [/bib_ref] , but have found their use as anti-tumour compounds. Their structures are shown in [bib_ref] Epothilones: mechanism of action and biologic activity, Goodin [/bib_ref]. They contain a methylthiazole group attached by an olefinic bond. They are active against breast cancer and other forms of cancer [bib_ref] Desoxyepothilone B is curative against human tumor xenografts that are refractive to..., Chou [/bib_ref]. They bind to and stabilize microtubules essential for DNA replication and cell division, even more so than taxol. One epothilone, ixebepilone (Ixempra), produced chemically from epothilone B and which targets microtubules, has been recently approved by FDA. By preventing the disassembly of microtubules, epothilones cause arrest of the tumour cell cycle at the GM2/M phase and induce apoptosis. The mechanism is similar to that of taxol but epothilones bind to tubulin at different binding sites and induce microtubule polymerization. Production of epothilone B by S. cellulosum is accompanied by the undesirable epothilone A. Production of epothilone B over A is favoured by adding sodium propionate to the medium. Epothilone polyketides are more water-soluble than taxol. The epothilone gene cluster has been cloned, sequenced, characterized and expressed in the faster growing S. coelicolor, resulting in the production of epothilones A and B [bib_ref] Cloning and heterologous expression of the epothilone gene cluster, Tang [/bib_ref].
## Anti-angiogenic agents
Angiogenesis is necessary for tumours to obtain oxygen and nutrients. Tumours actively secrete growth factors, which trigger angiogenesis. The concept of angiogenesis was established by Professor Judah Folkman [bib_ref] A review of Judah Folkman's remarkable achievements in biomedicine, Cao [/bib_ref]. He proposed that tumour growth depends on angiogenesis and proposed the use of angiogenesis inhibitors as anti-tumour agents, i.e. to target activated endothelial cells. He further proposed that the vascular endothelial growth factor (VEGF) is involved in angiogenesis and that it could be a target for anti-angiogenic drugs. Fumagillin, produced by Aspergillus fumigatis, was one of the first agents found to act as an anti-angiogenesis compound. Next to come along were its oxidation product ovalacin and the fumagillin analogue TNP470 (= AGM-1470) [bib_ref] Synthetic analogues of fumagillin that inhibit angiogenesis and suppress tumour growth, Ingber [/bib_ref] [bib_ref] Short enantioselective synthesis of (-)-ovalicin, a potent inhibitor of angiogenesis, using substrate-enhanced..., Corey [/bib_ref]. TNP470 binds to and inhibits type 2 methionine aminopeptidase (MetAP2) [bib_ref] Methionine aminopeptidase (type 2) is the common target for angiogenesis inhibitors AGM-1470..., Griffith [/bib_ref] [bib_ref] The anti-angiogenic agent fumagillin covalently binds and inhibits the methionine aminopeptidase, MetAP-2, Sin [/bib_ref]. This interferes with aminoterminal processing of methionine, which may lead to inactivation of enzymes essential for growth and migration of endothelial cells [bib_ref] AGM-1470, a potent angiogenesis inhibitor, prevents the entry of normal but not..., Antoine [/bib_ref] [bib_ref] Selective inhibition of amino-terminal methionine processing by TNP-470 and ovalicin in endothelial..., Turk [/bib_ref]. In animal models, TNP470 effectively treated many types of tumour and metastases [bib_ref] Inhibitory effect of angiogenesis inhibitor TNP-470 on tumor growth and metastasis of..., Yanase [/bib_ref] [bib_ref] Prevention of hepatic metastasis of human colon cancer by angiogenesis inhibitor TNP-470, Tanaka [/bib_ref] [bib_ref] Angiogenesis inhibitor TNP-470 inhibits human breast cancer osteolytic bone metastasis in nude..., Sasaki [/bib_ref]. The monoclonal antibody Avastin (bevacizumab) is a first-line remedy for metastatic coleorectal cancer and is an angiogenesis inhibitor. Antiangiogenic agents Pegaptanib (Macugen) and ranibizumab (Lucentis) were approved by FDA. Macugen is an aptomer of the VEGF and Lucentis is an anti-VEGF antibody. By the end of 2007, 23 anti-angiogenic drugs were in Phase III clinical trials and more than 30 were in Phase II. By 2008, 10 anti-angiogenesis drugs had been approved. Eight are used against cancer and two are employed for treatment of age-related macular degeneration. Anti-angiogenesis therapy is now known as one of the four major types of cancer treatment.
## Rapamycin (sirolimus) derivatives
This narrow spectrum polyketide antifungal agent [bib_ref] Rapamycin (AY-22, 989), a new antifungal antibiotic. I. Taxonomy of the producing..., Vezina [/bib_ref] attained stature not as an antibiotic but as a potent immunosuppressive agent used widely for organ transplantation. Rapamycin's mode of action is that of inhibiting the mTOR (mammalian target of rapamycin) phosphatidylinositol lipid kinase. Interestingly, it was also found to have anti-tumour activity [bib_ref] Rapamycin is active against B-precursor leukemia in vitro and in vivo, an..., Brown [/bib_ref] by interfering with angiogenesis [bib_ref] Rapamycin inhibits primary and metastatic tumor growth by antiangiogenesis: involvement of vascular..., Guba [/bib_ref] [bib_ref] Strange bedfellows in transplant drug therapy, Pray [/bib_ref] and inducing apoptosis. Rapamycin was the basis of chemical modification and these efforts yielded important products such as temsirolimus (CCI-779; (Torisel), everolimus and deforolimus (A23573). Temsirolimus, an mTOR protein kinase inhibitor, has been approved by FDA for renal cell carcinoma, while everolimus and deferolimus, are in clinical trials against cancer [bib_ref] Ready for a comeback of natural products in oncology, Bailly [/bib_ref].
## Statins
The statins inhibit de novo production of cholesterol in the liver, the major source of blood cholesterol, thus lowering cholesterol levels in humans. They are microbially produced enzyme inhibitors, inhibiting 3-hydroxy-3-methylcoenzyme A reductase, the regulatory and rate-limiting enzyme of cholesterol biosynthesis in liver. As a result, they are the leading group of pharmaceuticals in the world. One of the most useful statins in lovastatin (monocolin K, mevinolin) produced by Monascus ruber [bib_ref] Monacolin K, a new hypocholsterolemic agent produced by a Monascus species, Endo [/bib_ref] and Aspergillus terreus [bib_ref] Mevinolin. A highly potent competitive inhibitor of hydroxymethylglutaryl-coenzyme A reductase and a..., Alberts [/bib_ref]. Statins also have anti-cancer activity, i.e. they inhibit in vitro and in vivo growth of pancreatic tumours. They also sensitize tumours to cytostatic drugs such as gemcitabrine which is used for pancreatic cancer [bib_ref] Statins inhibit Akt/PKB signaling via P2X7 receptor in pancreatic cancer cells, Mistafa [/bib_ref]. It has been shown that there is over a 50% reduction in risk of metastatic or fatal prostate cancer among people taking statins and an 80% reduction in pancreatic cancer among people using statins for 4 years. It has been found that lovastatin has anti-tumour activity against Lewis Lung Carcinoma cells [bib_ref] The Monascus metabolite monacolin K reduces tumor progression and metastasis of Lewis..., Ho [/bib_ref].
## Searching for new agents
High-throughput screening of 16 000 compounds for selective inhibition of cancer stem cells surprisingly revealed salinomycin to be the best [bib_ref] Identification of selective inhibitors of cancer stem cells by high-throughput screening, Gupta [/bib_ref]. In fact, it was 100 times more active than taxol. Salinomycin is used as a coccidiostat in poultry and other livestock and as an agent increasing feed efficiency in ruminant animals. Also active were etoposide, abamectin and nigericin. Both salinomycin and nigericin are structurally related polyether potassium ionophores.
Genetic manipulation is one way to obtain novel antitumour agents. This is carried out by expressing biosynthetic gene clusters from anti-tumour pathways in other organiasms. This has resulted in the formation of some novel hydroxylated and glycosylated anti-tumour agents [bib_ref] Genetic manipulation of antitumor-agent biosynthesis to produce novel drugs, Salas [/bib_ref].
The concept of genome mining was developed after it was found that the S. coelicolor genome contained 22 gene clusters encoding production of secondary metabolites but only four known products. The technique is useful for identifying genetic units with potential for synthesizing new drugs and has yielded many new antibiotics and anti-tumour agents [bib_ref] Mining and engineering natural-product biosynthetic pathways, Wilkinson [/bib_ref] [bib_ref] Genomic mining -a concept for the discovery of new bioactive natural products, Gross [/bib_ref] [bib_ref] Strategies for the discovery of new natural products by genome mining, Zerikly [/bib_ref]. One anti-cancer compound developed by Thallion Pharmaceuticals is ECO-4601. It inhibits the Ras-mitogenactivated phosphokinase (MAPK) pathway and binds selectively to PBR (peripheral benzodiazapine receptor), which is overexpressed in many types of tumour. It is currently in clinical trials. As a result of genome mining with Micromonospora sp., a new anti-tumour drug was discovered [bib_ref] A genomics-guided approach for discovering and expressing cryptic metabolic pathways, Zazopoulos [/bib_ref]. The compound, ECO-04601, is a farnesylated dibenzodiazapene, which induces apoptosis.
A new antibiotic showing anti-tumour activity was isolated from a co-culture present in the drainage of an abandoned mine, where the prevalent nutrient conditions created an extreme environment [bib_ref] Glionitrin A, an antibiotic-antitumor metabolite derived from competitive interaction between abandoned mine..., Park [/bib_ref]. Such an environment results in defensive and offensive microbial interactions for survival of microbes. The two members of the co-culture were the bacterium Sphingomonas sp. strain KMK-001, and the fungus Aspergillus fumigatus strain KMC-90. A new diketopiperazine disulfide, glionitrin A, was isolated from the co-culture but was not detected in monoculture broths of KMK-001 or KMC-901. Glionitrin A is a (3S, 10aS) diketopiperazine disulfide containing a nitro aromatic ring. It displayed significant antibiotic activity against a series of microbes including methicillin-resistant Staphylococcus aureus. An in vitro MTT cytotoxicity assay revealed that it has potent submicromolar cytotoxic activity against four human cancer cell lines: HCT-116, A549, AGS and DU145.
## Plant anti-tumour products
Plants have been a useful source of approved anti-cancer agents [bib_ref] A review on plant-derived natural products and their analogs with anti-tumor activity, Dholwani [/bib_ref]. Since 1961, nine plantderived compounds have been approved for use as anticancer drugs in the USA [fig_ref] Table 2: Some approved plant-derived anti-tumour compounds [/fig_ref]. Various groups are described below.
## Alkaloids
Vinca monoterpene indole alkaloids, such as vinblastine and vincristine, orginate from the Madagascar periwinkle plant (Catharanthus roseus). Vinblastine is commonly used to treat cancers such as Hodgkin's lymphoma. Vinblastine and vincristine have important pharmacological activities but are synthetically challenging. Metabolic engineering of alkaloid biosynthesis can provide an efficient and environmentally friendly route to analogues of these pharmaceutically valuable natural products. The enzyme at the entry point of the pathway, strictosidine synthase, has a narrow substrate range and thus limits a pathway engineering approach. However, it has been demonstrated by that with a different expression system and screening method, it is possible to rapidly identify strictosidine synthase variants that accept tryptamine analogues not utilized by the wildtype enzyme. The variants are used in a stereoselective synthesis of b-carboline analogues and are assessed for biosynthetic competence within the terpene indole alkaloid pathway. These results present an opportunity to explore metabolic engineering of 'unnatural' product production in the plant periwinkle.
Madagascar periwinkle also produces serpentines, which have shown promise as anti-cancer agents. Other plant-produced compounds have shown pharmacological activities including anti-cancer activity but are too toxic for use in humans. However, researchers have been able to produce a range of halogenated alkaloids. This strategy could help expand the range of available drug candidates for cancer.
Camptothecin is a modified monoterpene indole alkaloid produced by certain plants (angiosperms) [bib_ref] Camptothecin and taxol: from discovery to clinic, Wall [/bib_ref]. It also is produced by the endophytic fungus, Entrophospora infrequens, from the plant Nathapodytes foetida. In view of the low concentration of camptothecin in tree roots and poor yield from chemical synthesis, the fungal fermentation is very promising for industrial production. Camptothecin is used for recurrent colon cancer and has unusual activity against lung, ovarian, and uterine cancer [bib_ref] Bioreactor studies on the endophytic fungus Entrophospora for the production of an..., Amna [/bib_ref]. Colon cancer is the second leading cause of cancer fatalities in the USA and the third most common cancer among US citizens. Camptothecin is known commercially as Camptosar and Campto and achieved sales of $1 billion in 2003 [bib_ref] Camptothecin, over four decades of surprising findings, Lorence [/bib_ref]. Camptothecin's water-soluble derivatives irinotecan and topotecan are also used clinically.
The cellular target of camptothecin is type I DNA topoisomerase. When patients become resistant to irenotecan, its use can be prolonged by combining it with the monoclonal antibody Erbitux (Cetuximab). Erbitux blocks a protein that stimulates tumour growth and the combination helps metastatic colorectal cancer patients express- Natural products -cancer 693 ing epidermal growth factor receptor (EGFR). This protein is expressed in 80% of advanced metastatic colorectal cancers. The drug combination reduces invasion of normal tissues by tumour cells and the spread of tumours to new areas.
Taxol (paclitaxel), a diterpene alkaloid, has been a very successful anti-tumour molecule. An outline of taxol biosynthesis is shown in [fig_ref] Figure 4: Outline of taxol biosynthesis [/fig_ref] [bib_ref] Genetic engineering of taxol biosynthetic genes in Saccharomyces cerevisiae, Dejong [/bib_ref]. It was originally discovered in plants but has also been found to be a fungal metabolite [bib_ref] Taxol and taxane production by Taxomyces andreanae, an endophytic fungus of Pacific..., Stierle [/bib_ref]. Fungi, such as Taxomyces adreanae, Pestalotiopsis microspora, Tubercularia sp. and Phyllosticta citricarpa, produce taxol [bib_ref] Taxol and taxane production by Taxomyces andreanae, an endophytic fungus of Pacific..., Stierle [/bib_ref] [bib_ref] Endophytic taxol-producing fungi from bald cypress, Li [/bib_ref] [bib_ref] Taxol from Tubercularia sp. Strain TF5, an endophytic fungus of Taxus mairei, Wang [/bib_ref]. Production by P. citricarpa is rather low, i.e. 265 mg l -1 [bib_ref] Taxol from Phyllosticta citricarpa, a leaf spot fungus of the angiosperm Citrus..., Kumaran [/bib_ref]. However, it is claimed that another fungus, Alternaria alternate var monosporus from the bark of Taxus yunanensis, after ultraviolet and nitrosoguanidine mutagenesis, can produce taxol at the high level of 227 mg l -1 [bib_ref] Screening the high-yield paclitaxel producing strain Alternaria alternate var monosporus, Duan [/bib_ref]. Originally isolated from the bark of the Pacific yew tree (Taxus brevifolia), taxol showed anti-tumour activity but it took six trees of 100 years of age to treat one cancer patient [bib_ref] Taxol (paclitaxel): mechanisms of action, Horwitz [/bib_ref]. Today, it is produced by plant cell culture or by semisynthesis from taxoids made by Taxus species. These species make more than 350 known taxoid compounds [bib_ref] The taxane diterpenoids, Baloglu [/bib_ref].
Early genetic engineering of Saccharomyces cerevisiae yielded no taxadiene (the taxol precursor) because too little of the intermediate, geranylgeranyl diphosphate, was formed. When the Taxus canadensis geranylgeranyl diphosphate synthase gene was introduced into S. cerevisiae, 1 mg l -1 of taxadiene was obtained [bib_ref] Genetic engineering of taxol biosynthetic genes in Saccharomyces cerevisiae, Dejong [/bib_ref]. More recent metabolic engineering has yielded a S. cerevisiae strain producing over 8 mg l -1 taxadiene and 33 mg l -1 of geranyl geraniol.
The use of cells of the plant Taxus chinensis to produce taxol became the industrial means to make the compound. The addition of methyl jasmonate, a plant signal transducer, increased production from 28 to 110 mg l -1 [bib_ref] Methyl jasmonate-induced overproduction of paclitaxel and baccatin III in Taxus cell suspension..., Yukimune [/bib_ref]. The optimum temperature for growth of T. chinensis is 24°C and that for taxol synthesis is 29°C. Shifting from 24°C to 29°C at 14 days gave 137 mg l -1 at 21 days [bib_ref] Intermittent maltose feeding enhances paclitaxel production in suspension culture of Taxus chinensis..., Choi [/bib_ref]. There is a 6-week process yielding 153 mg l -1 with T. chinensis. Mixed cultures of Taxus plant cells and taxol-producing endohytic fungi are under investigation .
Taxol had sales of $1.6 billion in 2005. It is approved for breast and ovarian cancer and acts by blocking depolymerization of microtubules. In addition, taxol promotes tubulin polymerization and inhibits rapidly dividing mammalian cancer cells [bib_ref] Taxol: an antimitotic agent with a new mechanism of action, Manfredi [/bib_ref]. Taxanes and camptothecins alone accounted for approximately onethird of the global anti-cancer market in 2002, with a market value of over $2.75 billion. An analogue of taxol, docetaxel (Taxotere), had sales of $3 billion in 2009 [bib_ref] More than a supplier, Thayer [/bib_ref].
Taxol also has antifungal activity by the same microtubule mechanism especially against oomycetes (Strobel and [bib_ref] Endophytic microbes embody pharmaceutical potential, Strobel [/bib_ref] [bib_ref] Rainforest endophytes and bioactive products, Strobel [/bib_ref]. Oomycetes are water moulds exemplified by plant pathogens, such as Phytophthora, Pythium and Aphanomyces.
## Etoposide and teniposide
These two compounds were derived as semisynthetic derivatives of podophyllotoxin, an antimitotic metabolite of the roots of the mayapple plant, Podophyllum peltatam [bib_ref] Back to basics: how natural products can provide the basis for new..., Deorukhkar [/bib_ref] [bib_ref] Natural compounds for cancer treatment and prevention, Nobili [/bib_ref]. The mayapple plant is an old herbal remedy. Etoposide is a topoisomerase II inhibitor. This essential enzyme is involved in eukaryotic cell growth by regulating levels of DNA supercoiling [bib_ref] Substituents on etoposide that interact with human topoisomerase IIa in the binary..., Bender [/bib_ref]. Etoposide was approved for lung cancer, choriocarcinoma, ovarian and testicular cancer, lymphoma, and acute myeloid leukaemia. Teniposide was approved for tumours of the central nervous system, malignant lymphoma and bladder cancer.
## Other compounds
The naphthoquinone pigment shikonin, a herbal medicine remedy, is produced by cell culture of the plant Lithospermum erythrorhizon, mainly for cosmetic use. Unexpectedly, shikonin and two derivatives were found to inhibit tumour growth in mice bearing Lewis Lung Carcinoma [bib_ref] Shikonin, acetylshikonin, and isobutyroylshikonin inhibit VEGF-induced angiogenesis and suppress tumor growth in..., Lee [/bib_ref]. Other plant natural products such as the isoflavone genistine, indole-3-carbinol (I3C), 3,3′diindolemethane, curcumin (-)-epigallocatechin-3-gallate, resveratrol and lycopene are known to inhibit the growth of cancer cells [bib_ref] Cellular signalling perturbation by natural products, Sarkar [/bib_ref]. These natural compounds appear to act by interference in multiple cellular signalling pathways, activating cell death signals, and bringing on apoptosis of cancer cells without negatively affecting normal cells.
## Marine anti-tumour products
Although the ocean represents the centre of biological diversity with 34 of 37 phyla of life represented (compared with only 17 on land), prospecting marine resources for biotechnological use, particularly in drug discovery, is a relatively recent activity. Marine microorganisms encompass a complex and diverse assemblage of microscopic life forms, of which it is estimated that only 1% have been cultured or identified. Coral reefs and other highly diverse ecosystems, such as mangroves and sea grass, have been targeted for bioprospecting because they host a high level of biodiversity and are often characterized by intense competition for space, leading to chemical warfare among sessile organisms. Marine sponges produce numerous bioactive compounds with promising pharmaceutical properties. Cytarabine (Cytostar) used for non-Hodgkin's lymphoma was originally derived from a sponge [bib_ref] Oceans: medicine chests of the future?, Rayl [/bib_ref]. Sponges have been frequently hypothesized to contain compounds of bacterial origin and the bacterial symbionts have long been suspected to be the true producers of many drug candidates. Marine products with anti-tumour activity are listed in [fig_ref] Table 3: Marine products with anti-tumour activity [/fig_ref].
Scientists have recently identified an uncultured Pseudomonas sp. symbiont as the most likely producer of the defensive anti-tumour polyketide pederin in Paederus fuscipes beetles by cloning the putative biosynthesis genes. Closely related genes have also been isolated from the highly complex metagenome of the marine sponge Theonella swinhoei, which is the source of the onnamides and theopederins, a group of polyketides that structurally resemble pederin. Sequence features of the isolated genes clearly indicate that they may belong to a prokaryotic genome and are responsible for the biosynthesis of almost the entire portion of the polyketide structure that is correlated with anti-tumour activity. Besides providing further proof for the role of the related beetle symbiont-derived genes, these findings raise intriguing ecological and evolutionary questions and have important general implications for the sustainable production of otherwise inaccessible marine drugs by using biotechnological strategies [bib_ref] Antitumor polyketide biosynthesis by an uncultivated bacterial symbiont of the marine sponge..., Piel [/bib_ref].
Sixty-eight per cent of the pharmaceutically useful marine natural products employed are for cancer and the rest are used for inflammation, pain, asthma and Alzheimer's disease. Global sales of marine biotechnology products including anti-cancer compounds exceeded $3.2 billion in 2007. Since 2007, the marine alkaloid trabectedin (Yondelis) was approved by FDA [bib_ref] Ready for a comeback of natural products in oncology, Bailly [/bib_ref]. Another anti-tumour compound, aplidine, a cyclic peptide, has Orphan Drug Status in Europe for acute lymphocytic leukaemia and is in Phase II clinical trials in the USA. Ecteinascidin 743 (ET 743) is in Phase III trials against sarcoma. A major problem in this field is that less than 1% of the commensal microbiotic consorta of marine invertebrates are culturable at this time.
Curacin A, obtained from a marine cyanobacterium Lyngbya majuscula isolated in Curacao, showed potent anti-tumour activity [bib_ref] Characterization of the interaction of the marine cyanobacterial natural product curacin A..., Blokhin [/bib_ref]. Other antitumour agents derived from marine sources include eleutherobin, discodermolide, bryostatins, dolastatins and cephalostatins.
The symbionts of marine invertebrate animals continue to reveal interesting natural products 2007). Variants of the toxic dolastin from the sea hare Dolabella auricalaria seem promising against cancer. These include soblidofin (T2F 1027), which completed Phase II against soft tissue sarcoma, and synthadotin (tasidotin, 1LX 651), which is at the same clinical stage against melanoma, prostate and non-small cell lung cancers. These are thought to be produced by cyanobacteria sequestered by the marine invertebrates in their diet.
The marine environment continues to offer new opportunities for the discovery of anti-tumour agents [bib_ref] Marine actinomycete diversity and natural product discovery, Jensen [/bib_ref]. The new actinomycete genus Salinispora and its two species, S. tropica and S. arenicola, have been isolated around the world. These require seawater for growth. Salinispora tropica makes a new bicyclic g-lactone b-lactam called salinosporamide A, which is a proteasome inhibitor [bib_ref] Structure-activity relationship studies of salinosporamide A (NPI-0052), a novel marine derived proteasome..., Macherla [/bib_ref] in Phase II clinical trials against multiple myeloma and mantle cell lymphoma. Also the genus Marinophilus contains species that produce novel polyenes, which have no antifungal activity but display potent anti-tumour activity.
## Final comments
How can we come up with the anti-tumour compounds of the future? High-throughput screening and combinatorial chemistry have not provided the numbers of high-quality leads that were anticipated. Combinatorial chemistry mainly yields minor modifications of present day drugs and absolutely requires new scaffolds, such as natural products, on which to build. Although comparative genomics is capable of disclosing new targets for drugs, the number of targets is so large that it requires tremendous investments of time and money to set up all the screens necessary to exploit this resource. This can only be handled by high-throughput screening methodology which demands libraries of millions of chemical entities. It is clear that the future success of the pharmaceutical industry depends not only on high-throughput screening and combinatorial chemistry, but on the combining of complementary technologies, such as natural product discovery, genomics, proteomics, metabolomics, metagenomics, structure-function drug design, semi-synthesis, recombinant DNA methodology, genome mining and combinatorial biosynthesis. In addition, 'intelligent screening' methods, robotic separation with structural analysis, metabolic engineering and synthetic biology offer exciting technologies for new natural product drug discovery and the future development of anti-tumour compounds.
[fig] Figure 2, Figure 3: Abbreviated Structures of epothilones. Reprinted fromGoodin and colleagues (2004) with permission. [/fig]
[fig] Figure 4: Outline of taxol biosynthesis. Reprinted from DeJong and colleagues (2005) with permission. [/fig]
[table] Table 1: Some microbial anti-tumour compounds.Fig. 1.Structures of some anti-tumour agents with clinical application. Reprinted fromSalas and Mendez (1998) with permission. [/table]
[table] Table 2: Some approved plant-derived anti-tumour compounds. [/table]
[table] Table 3: Marine products with anti-tumour activity. [/table]
|
A comparative assessment of antiproliferative properties of resveratrol and ethanol leaf extract of Anogeissus leiocarpus (DC) Guill and Perr against HepG2 hepatocarcinoma cells
Background: Epidemiological and experimental evidences have shown cancer as a leading cause of death worldwide. Although the folklore use of plants as a reliable source of health-restoring principles is well-documented, the search for more of such plants that are active against diseases, such as cancer, continues. We report here a laboratory-based evidence of the relevance of an ethanol leaf extract of Anogeissus leiocarpus (A2L) in comparison with resveratrol, a natural polyphenol, in cancer therapy.Methods:The quantitative assessment of flavonoid and phenolic contents involved quercetin and gallic acid as standards, respectively were determined using spectrophotometry. Cytotoxicity was determined fluorometrically using propidium-iodide-staining method. Antioxidant status, adenosine triphosphate (ATP) levels, caspase activities and mitochondrial integrity were assessed using fluorometry/luminometry.Results:The antioxidant assay demonstrated that A2L possesses a strong antioxidant capacity as compared with the reference compounds, ascorbic acid and butylated hydroxytoluene. This is further buttressed by the significantly high level of phenolics obtained in the quantitative assessment of the extract. A 72-h post-treatment examination indicated that both A2L and resveratrol modulate the proliferation of HepG2 liver carcinoma cells in a time-and concentrationdependent manner. Determination of the total nuclei area, propidium-iodide negative and positive nuclei areas all further buttress the modulation of cell proliferation by A2L and resveratrol with the indication that the observed cell death is due to apoptosis and necrosis at lower and higher concentrations of treatments respectively. At lower concentrations (0.39-3. 13 μg/mL), resveratrol possesses higher tendencies to activate caspases 3 and 7. Bioenergetically, both resveratrol and A2L do not adversely affect the cells at lower concentrations (0.39-6.25 μg/mL for resveratrol and 12.5-100.0 μg/mL for A2L) except at higher concentrations (12.5-25.0 μg/mL for resveratrol and 200-800 μg/mL for A2L) which are more pronounced in A2L-treated cells. Furthermore, the antioxidant status of HepG2 cells is not perturbed by resveratrol as compared with A2L. Assessment of 24-h post-treatment mitochondrial function shows that resveratrol is not mitotoxic as compared with A2L which exhibits mitotoxicity at its highest concentration.
## (continued from previous page)
Conclusions: Taken together, findings from this study showed that A2L possesses strong antiproliferative activity and its prospect in the management of hepatocellular carcinoma deserves further investigation.
Keywords: Free radicals, Resveratrol, Anogeissus leiocarpus, HepG2 cells, Cell death, Cell proliferation, Morphological changes
# Background
The current statistics reveal that cancer is now a leading cause of death worldwide, and liver cancer is one of the deadliest forms. Liver cancer has been reported to be the second most common cause of death from cancer globally after lung cancer [bib_ref] cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN..., Ferlay [/bib_ref]. Increasingly new cases of liver cancer were diagnosed, from 626,162 in 2002 to 782,451 in 2012. This figure is projected to increase to 1,341,344 cases by 2035 [bib_ref] cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN..., Ferlay [/bib_ref].
There has been an increased search for anticancer medicinal plant products based on the fact that surgical and traditional therapeutic approaches (chemotherapy and radiation) are, at present, unable to control most cancer types [bib_ref] Cell survival and apoptosis signaling as therapeutic target for cancer: marine bioactive..., Kalimuthu [/bib_ref]. Furthermore, considering the concomitant toxic effects associated with the use of synthetic chemicals in cancer therapy, scientific research efforts are being geared toward naturally-occurring plant sources which are therapeutically effective, culturally acceptable and economically accessible [bib_ref] Chemoprevention of aflatoxin B 1 -induced genotoxicity and hepatic oxidative damage in..., Farombi [/bib_ref].
Anogeissus leiocarpus (DC) Guill and Perr (English/ common name: Axlewood tree; Yoruba name: Ayin) is one of the medicinal plants contained in the Nigeria's diverse flora . It has varied applications in Nigerian traditional medicine mostly in the treatment of respiratory diseases, ascaricide, gonorrhoea, general body pain, blood clots. [bib_ref] An ethnobotanical survey of indigenous flora for treating tuberculosis and other respiratory..., Mann [/bib_ref]. The result of an oral acute toxicity study revealed no death with doses up to 3200 mg/kg body weight in an animal model. Furthermore, histopathological lesions were not observed in all the organs except the lungs, which showed congestion, oedema and bronchitis. These results suggested that the leaf extract of A. leiocarpus could be used with some degree of safety especially by oral route [bib_ref] Acute toxicity effects of the aqueous leaf extract of Anogeissus leiocarpus in..., Agaie [/bib_ref].
Resveratrol [bib_ref] Acute toxicity effects of the aqueous leaf extract of Anogeissus leiocarpus in..., Agaie [/bib_ref] ,4′-Trihydroxystilbene) is a natural polyphenol which was isolated from the root of Veratrum grandiflorum by Takaoka with its chemical structure characterised in 1940 [bib_ref] Of the phenolic substances of white hellebore (Veratrum grandiflorum Loes. Fil.), Takaoka [/bib_ref]. Studies have shown that resveratrol can exert antitumour effects at the stages of cancer initiation, promotion and progression [bib_ref] Differential effects on growth, cell cycle arrest, and induction of apoptosis by..., Hsieh [/bib_ref]. Moreover, resveratrol has also received considerable attention due to its potential chemopreventive and chemotherapeutic properties [bib_ref] Potentiation of resveratrol-induced apoptosis by matrine in human hepatoma HepG2 cells, Ou [/bib_ref]. This study, therefore, evaluated the in vitro cell modulatory properties of ethanol leaf extract of A. leiocarpus (A2L) as compared with resveratrol in HepG2 liver carcinoma cell line.
# Methods
Collection of leaves of A. leiocarpus and preparation of its ethanol extract A. leiocarpus leaves were collected from Lokoja, Nigeria and authenticated at the Department of Botany, University of Ibadan with the reference number: UIH 22402. The sample leaves were washed with distilled water (dH 2 O) to remove any contaminants, air-dried under shade until it attained a constant weight, grinded to powder, sieved, packed into polythene bags and stored at 4°C. Four hundred grams (400 g) of the powered plant part was macerated in 70% ethanol (1600 mL) for 72 h with intermittent stirring/shaking [bib_ref] In vitro evaluation of the antioxidant potential, phenolic and flavonoid contents of..., Olugbami [/bib_ref]. At the end of the extraction, the extract was filtered using Whatman filter paper No.1 (Whatman Ltd., England) to remove all unextractable matters, including cellular materials and other constituents that are insoluble in the extraction solvent. The filtrate was concentrated using a rotary evaporator (RE-52A, Shanghai Ya Rong Biochemistry Instrument Factory, Shanghai) under reduced pressure (in order to speed up the process) at 40°C and stored at 4°C until when needed. The percentage yield of the extraction was 14.17% w/w.
## Determination of total phenolic content
The amount of total phenolics in the plant extract was determined with the Folin-Ciocalteau reagent using the method of Spanos and Wrolstad [bib_ref] Influence of processing and storage on the phenolic composition of Thompson seedless..., Spanos [/bib_ref] with some modifications an in Olugbami et al. [bib_ref] In vitro free radical scavenging and antioxidant properties of ethanol extract of..., Olugbami [/bib_ref]. Briefly, 2.5 mL of 1/ 10 dilution of Folin-Ciocalteau's reagent and 2 mL of Na 2 CO 3 (7.5% w/v) were added to 0.50 mL of the extract (800 μg/mL) and incubated at 45°C for 15 mins. The absorbance of was measured at 765 nm using UV/VIS spectrometer T70. Results are expressed as micrograms of gallic acid equivalents per milligram of dry weight (μg GAE/mg) of extract.
## Determination of total flavonoid content
This was carried out by using the method described by Nickavar and colleagues [bib_ref] Comparison of the components of the essential oils from leaves and fruits..., Nickavar [/bib_ref]. Briefly, 2.5 mL of the extract (800 μg/mL) was mixed with 2.5 mL AlCl 3 reagent in 90% ethanol and allowed to stand for 40 mins at room temperature. After that, the absorbance of the mixture was measured at 415 nm using UV/VIS spectrometer T70. The blank was made up of 2.5 mL of 90% ethanol plus sample solution (2.5 mL). The TFC which is expressed as micrograms of quercetin equivalents per milligram (μg QE/mg) of extract was determined on the basis of the linear calibration curve of quercetin (absorbance versus quercetin concentration).
## Cell line and reagents
## Cell counting, image acquisition and analysis
Cells were harvested and re-suspended in medium to ensure a single cell suspension and then counted on a haemocytometer based on trypan blue dye exclusion method (0.4% trypan blue; Sigma-Aldrich, St Louis, MO) by adding 10 μL of trypan blue solution to 10 μL of cell suspension (1:1 v/v). ImageXpress® Micro Widefield High Content Screening System and its associated MetaXpress® High Content Image Acquisition and Analysis Software interface were utilised for image acquisition and analysis. After dye addition and incubation (1 h), images were captured with two software predetermined excitation/emission filter sets; 380/535 nm for Hoechst and 555/645 nm for PI. Transmitted light and fluorescence images from each well were acquired using a 20× and 4× dry objective lenses respectively. Under these settings and by plating HepG2 cells at the predetermined optimum cell densities indicated above, the system was able to capture from~100 to 400 cells per well and the images were subsequently analysed using the MetaXpress® Analysis Software. To define nuclei as individual units or regions of interest (ROIs), preprocessing filters and intensity thresholds were applied for image segmentation. Segmented images were subjected to data classification by the use of the MetaX-press® Analysis Software. The percentage of dead cells was calculated from the total number of ROIs per well. Cell nuclei emitting fluorescence signal from both Hoechst and PI (fluorescence co-localisation) were considered as dead cells, while cells emitting only Hoechst signal were counted as live cells.
## Cytotoxicity determination
Resveratrol (100 mg) was prepared as 20 mg/mL by adding 5 mL of 100% DMSO and then sonicated to ensure total dissolution of the solute. While the extract was prepared in 0.5% DMSO, sonicated and centrifuged at 13500 rpm with the supernatant preserved for subsequent use. For the assays, resveratrol and extract were further diluted to appropriate concentrations in complete medium. In both cases, the final DMSO concentration did not exceed 0.25%. Having obtained an optimum DMSO concentration and cell density, cells were prepared and then treated with resveratrol (0.39, 0.78, 1.56, 3.13, 6.25, 12.5, and 25 μg/mL) and A2L , 400, and 800 μg/mL). The treated cells were incubated for 72 h with the effects of treatment being monitored at 24 h-intervals. At each interval, the plates were stained and then imaged with an ImageXpress MicroXL high-content fluorescence microscope.
## Mitochondrial toxicity testing
The Mitochondrial ToxGlo™ Assay used here is a cell-based assay that employs multiplexed chemistry for predicting potential mitochondrial dysfunction as a result of xenobiotic exposure. The assay is based on the differential measurement of biomarkers associated with changes in cell membrane integrity and cellular ATP levels relative to vehicle-treated control cells. Ten microlitres (10 μL) of culture medium containing approximately 1500 cells is plated on opaque-walled, black tissue culture plates compatible with fluorescent and luminescent measurements (clear bottoms). Being attachment-dependent, cells were incubated for 24 h to adhere to the substrate of the plate. The cells were then treated with 10 μL of sample and incubated for 24 h. Using an electronic multichannel pipette, 5 μL of 5X Cytotoxicity Reagent was dispensed to each well of the plate and mixed briefly (1 min) by orbital shaking (1300 rpm) to ensure Reagent/sample homogeneity and to mitigate possible Reagent/sample partitioning. The plate was then incubated at 37°C for 30 mins and fluorescence was measured with rhodamine 110 filter set at 485 nm Ex /525 nm Em using Flexstation 3 microplate reader (Molecular Devices, Sunnyvale, CA). Thereafter, the assay plate was equilibrated to room temperature (10 mins) and 25 μL of ATP Detection Reagent was added to each well. The plate was mixed by orbital shaking (1300 rpm) for 3 mins and luminescence was measured by using Wallac microplate reader.
## Caspase 3/7-activity
Caspase 3/7-activity was assessed with Caspase-Glo 3/7 luminescent Assay reagents. Cells were plated at 1500 per well in a volume of 12.5 μL culture medium in white 384-multiwell plates and allowed to attach for 24 h. After treatments (vehicle or test substances) in a volume of 12.5 μL culture medium, plates were subsequently incubated for 24 h. Then 25 μL of Caspase-Glo 3/7 Reagent was added to each well containing 25 μL of blank, negative control cells or treated cells in culture medium. Contents of wells were mixed using a plate shaker at 1300-1500 rpm for 30 s and the plates were incubated at room temperature for 1 h before measuring luminescence on Analyst GT multimode reader (Molecular Devices Corporation, Sunnyvale, CA, USA).
## Determination of adenosine triphosphate (atp) levels
ATP levels were measured using CellTiter-Glo Luminescent Cell Viability Assay. Cells were plated at 3000 per well in triplicate in white 384-multiwell plates and allowed to attach for 24 h. After treatments (vehicle or test substances) in a volume of 12.5 μL culture medium, the plates were subsequently incubated for 24 h and then a volume of CellTiter-Glo reagent equal to the content of each well was added. After 10 mins, luminescence was read using Wallac plate reader (Perkin-Elmer, USA), and signal intensity was calculated relative to inplate DMSO control wells.
## Determination of glutathione levels (antioxidant status)
Reduced-oxidised glutathione levels were determined by using the luminescence-based GSH/GSSG-Glo™ Assay reagents. Cells were plated at 3000 per well in a volume of 25 μL culture medium in white 384-multiwell plates and allowed to attach for 24 h. After treatments (vehicle or test substances) in a volume of 25 μL culture medium, plates were subsequently incubated for 24 h, cell culture medium/treatment was removed and plates were treated according to manufacturer's instructions to determine the levels of total and oxidised glutathione. After 15 mins of incubation to stabilise the signal, luminescence was read using Wallac plate reader (Perkin-Elmer, USA). GSH/GSSG ratios are calculated directly from relative luminescent unit (RLU) measurements.
# Statistical analysis
For statistical analysis, data were analysed using Sig-maPlot 11.0 and GraphPad Prism 5. The results were expressed as mean ± SD (standard deviation) and the median inhibitory/cytotoxic concentration (I/CC 50 ) values were obtained from the linear regression plots. One-way analysis of variance (ANOVA) was used to assess differences between means (student's t test, where applicable); if significant differences were found (p < 0.05, 0.01, or 0.001), Tukey-Kramer multiple comparison test was considered for post-test.
# Results
Quantitative phytochemical analysis shows the significantly higher content of phenolics than flavonoids in A2L
A comparison of the total phenolic and flavonoid contents based on gallic acid and quercetin as the standard compounds, respectively indicates that A2L is highly rich in phenolic contents (605.6075 ± 0.6182 μg GAE/ mg extract) as compared with its flavonoid contents (119.4792 ± 3.1302 μg QE/mg extract).
## Resveratrol displays mitogenic activity while a2l shows antiproliferative activity
Assessment of the proliferative activities of resveratrol and A2L in HepG2 cells for 72 h shows, on one hand, that resveratrol is mitogenic at lower concentrations throughout the period of treatment. On the other hand, A2L is consistent in its antiproliferative activity. For both resveratrol and A2L, cell proliferation seems not remarkably affected at 24 h after treatment. Overall, both resveratrol and A2L are antiproliferative at high concentrations being more pronounced with A2L which significantly (p < 0.05; 0.01; 0.001) displays a time-and dose-dependent antiproliferative activity. As expected, the negative control (DMSOtreated cells) maintained a similar proliferation pattern at the various concentrations within the same time of exposure .
## Resveratrol and a2l could not induce the death of hepg2 cells at lower concentrations
Assessment of percentage dead cell based on the number of cells that are positive for propidium iodide indicates that resveratrol and A2L could not significantly (p > 0.05) induce HepG2 cell death at lower concentrations for the 72 h of exposure. One sharp difference that can be noticed is that A2L shows higher tendencies to be death-inducing in a time-dependent manner than resveratrol. At the highest concentration, A2L maintains a significantly (p < 0.001) strong death-inducing ability from 24 h through 72 h of exposure .
Total, positive, and negative nuclear sizes are affected in a time-and concentration-dependent manner
The total nuclear area which indicates the overall combination of the area occupied by the genetic materials of both dead and live cells, followed the same trend with cell count . In the case of the negative nuclei total area, a parameter that measures the genetic/nuclear size of live cells, it shows no changes at lower concentrations but a significant (p < 0.01) increase at higher concentrations most especially at 72 h post-treatment . Lastly, the positive nuclei total area, which indicates the nuclear/genetic size of dead cells, shows a gradual increase in a time-and dose-dependent manner [fig_ref] Figure 5: Determination of positive nuclei total area [/fig_ref].
## Both resveratrol and a2l do not possess mitotoxic potential
Assessment of the influence of both resveratrol and A2L on possible mitotoxicity shows that both are unable to induce a depletion in the levels of ATP across the various concentrations of treatment considered in this study. The only exception being A2L which significantly depleted the ATP levels at its highest concentration [fig_ref] Figure 6: Assessment of mitochondrial toxicity [/fig_ref].
## Hepg2 cell death induced by both resveratrol and a2l involves the activation of executioner caspases
Across the various concentrations of treatment, resveratrol and A2L demonstrated the induction of the apoptosis-inducing caspases 3 and 7. For both treatments, activation of these caspases are more pronounced at the lower concentrations compare to the higher concentrations. As the concentrations increase, resveratrol shows a more drastic reduction in the levels of these caspases as compared with the minimal change recorded for A2L .
## Depletion in atp levels seems to account principally for the cell death induced by a2l
Both resveratrol and A2L could not significantly (p > 0.05) affect the ATP levels at lower concentrations, but A2L significantly (p < 0.05) exhibited ATP-depletion capacity at higher concentrations .
## Resveratrol and a2l are antioxidant boosters for hepg2 cells
The antioxidant status of HepG2 cells was unaffected significantly (P > 0.05) by resveratrol and A2L as compared with the negative control. The only exception being that A2L demonstrated higher reducing potential than resveratrol throughout all the concentrations of treatment ultimately culminating in a significant (p < 0.01) decrease in the antioxidant status at the highest concentration of A2L . Determination of total cells. Numerical changes associated with treatment of HepG2 cells with resveratrol (Resveratrol) or extract (A2L) followed up for three days. *p < 0.05; **p < 0.01; # p < 0.001 Quantification of treatment-induced cell death. Concentrationand time-dependent assessment of cytotoxicity resulting from treatment of HepG2 cells with resveratrol or A2L followed up for three days. *p < 0.05; **p < 0.01; # p < 0.001
## Resveratrol and a2l induced morphological changes in hepg2 cells
Morphological assessment of HepG2 cells shows a more pronounced effect induced on the cells by A2L most especially at its highest concentration as compared with resveratrol .
# Discussion
The high morbidity and mortality rates associated with hepatocellular carcinoma (HCC) has been attributed principally to its chemoresistance with proven resistance to chemotherapeutic agents, such as adriamycin (ADM), cisplatin, 5-fluorouracil (5-FU) and doxorubicin [bib_ref] Evaluation of anti-cancer potential of capsaicin-loaded trimethyl chitosan-based nanoparticles in HepG2 hepatocarcinoma..., Elkholi [/bib_ref] [bib_ref] LAPTM4B: a novel cancer-associated gene motivates multidrug resistance through efflux and activating..., Li [/bib_ref]. It is known that HCC is a major public health problem in many parts of the world, with high incident areas in East Asia and sub-Saharan Africa [bib_ref] Global cancer statistics in the year 2000, Parkin [/bib_ref]. Our efforts in this work were channeled towards application of a plantderived natural product in the management of HCC.
Various reports have shown that free radicals/reactive oxygen species (ROS), such as nitric oxide, hydroxyl radical, superoxide anion, are generated from both endogenous and exogenous sources. These products are generally reactive with cellular macromolecules, such as lipids, proteins, deoxyribonucleic acid (DNA). resulting in lipid peroxidation, degradation/denaturation of proteins, oxidation of DNA, respectively. These are the molecular basis of many diseases including cancer such as HCC [bib_ref] Glutathione in cancer cell death, Ortega [/bib_ref] [bib_ref] Oxidative stress, nutrition and health. Experimental strategies for optimization of nutritional antioxidant..., Halliwell [/bib_ref]. Findings from the present study demonstrated that the extract (A2L) used in this work possesses a very strong free radical scavenging activity as compared with ascorbic acid, the reference compound. This observed activity of A2L may be due to its high total phenolic content demonstrated in the quantitative phytochemical analysis. Undoubtedly, the flavonoid Determination of nuclei total area. Changes in total nuclear size associated with treatment of HepG2 cells with resveratrol or A2L followed up for three days. *p < 0.05; **p < 0.01; # p < 0.001 Determination of negative nuclei total area. Changes in negative nuclear size associated with treatment of HepG2 cells with resveratrol or A2L followed up for three days. *p < 0.05; **p < 0.01 content, although quite minimal, may also be a contributing factor. Plant phenolics and flavonoids have been reported to exhibit efficient radical scavenging, metal chelating and antioxidant activities [bib_ref] Antioxidant properties of several tropical fruits: a comparative study, Lim [/bib_ref] [bib_ref] Antioxidant properties and composition of aqueous extracts from Mentha species, hybrids, varieties,..., Dorman [/bib_ref]. The antiproliferative activity of resveratrol in HepG2 cells has been previously reported and here we confirm the same order except that its lower concentrations displayed mitogenic tendencies throughout the 72 h of exposure to treatment. In like manner, A2L almost demonstrated the same trend as resveratrol except that it exhibited greater antiproliferative activity in a time-and concentration-dependent manner [bib_ref] Resveratrol exerts its antiproliferative effect on HepG2 hepatocellular carcinoma cells, by inducing..., Notas [/bib_ref] [bib_ref] Resveratrol inhibits invasion and metastasis of hepatocellular carcinoma cells, Yu [/bib_ref]. The values obtained for total nuclear, positive nuclear, and negative nuclear areas all corroborated the antiproliferative, apoptotic and necrotic effects of treatments depending on the concentrations of either resveratrol or A2L. Antiproliferative activity is indicated by the decrease in total and negative nuclei areas, while the increase in positive nuclei area implies necrotic cell-death [bib_ref] Time-and concentration-dependent effects of resveratrol in HL-60 and HepG2 cells, Stervbo [/bib_ref] [bib_ref] Anticancer and apoptotic activities of oleanolic acid are mediated through cell cycle..., Zhu [/bib_ref] [bib_ref] Atomic force microscopy correlates antimetastatic potentials of HepG2 cell line with its..., Olugbami [/bib_ref].
In order to gain insights into the antiproliferative activities of both resveratrol and A2L, we assessed mitochondrial function and found that ATP synthesis was not significantly (p > 0.05) affected most especially at the lower concentrations, but became pronounced at higher concentrations. This is in line with a previous study [bib_ref] Resveratrol plus ethanol counteract the ethanol-induced impairment of energy metabolism: 31P NMR..., Gallis [/bib_ref]. Previous reports have indicated that resveratrol can induce mitochondrial biogenesis in HepG2 cells and may clearly explain the results we obtained in this present study [bib_ref] Resveratrol induces hepatic mitochondrial biogenesis through the sequential activation of nitric oxide..., Kim [/bib_ref] [bib_ref] Resveratrol improves health and survival of mice on a high-calorie diet, Baur [/bib_ref]. Although, this activity seems to be concentrationdependent as, at higher concentrations, both resveratrol and A2L were able to significantly (p < 0.05) reduce the ATP levels with A2L being potent (p < 0.01). The resultant effect of this disruption in ATP levels is the cell death induced at these concentrations. In accordance with earlier reports, we observed that resveratrol maintained the antioxidant status of HepG2 cells throughout the concentrations considered in this study. This activity has been reported to be due to the capability of resveratrol to induce an increase in the activities of catalase (CAT), superoxide dismutase (SOD), GSH peroxidase (GPx), NADPH quinone oxidoreductase, and GSH-S-transferase with an attendant increase in intracellular glutathione (GSH) level [bib_ref] Resveratrol protects HepG2 and chang liver cells from oxidative stress, Nah [/bib_ref] [bib_ref] Resveratrol and liver: a systematic review, Faghihzadeh [/bib_ref]. Results recorded on A2L indicated that the extract induced similar effect with resveratrol on the antioxidant status of HepG2 cells, with the exception of the highest concentration in which the extract significantly depleted the antioxidant level of the cells.
Resveratrol has been noted to induce apoptosis in cancer cells as it has been demonstrated in this study by the upregulation of the activities of effector caspases-3 and -7. For resveratrol-treated HepG2 cells, the trend of caspase activities obtained here across the various concentrations implies that the level of apoptotic cell death attained a threshold point after which other forms of cell death, most probably necrosis, becomes predominant. This point is evident from the decline in caspase activity. As for the A2L-treated cells, apoptotic level seems to have been well-maintained except at the highest concentration in which there was a remarkable reduction in the caspase activities [bib_ref] Potentiation of resveratrol-induced apoptosis by matrine in human hepatoma HepG2 cells, Ou [/bib_ref] [bib_ref] Resveratrol enhances p53 acetylation and apoptosis in prostate cancer by inhibiting MTA1/NuRD..., Kai [/bib_ref].
# Conclusion
Our findings validated the antiproliferative activities of resveratrol in HepG2 hepatocellular carcinoma cells. Here, for the first time we have been able to Caspase activation levels. This shows the trend of caspase activities induced by treatment with either resveratrol or A2L Effects of treatment on ATP levels. Variations in ATP status of HepG2 cells treated with resveratrol or A2L. *p < 0.05; **p < 0.01 Effects of treatment on anti-oxidant status. Anti-oxidant status (GSH/GSSG ratio) of HepG2 cells treated with resveratrol or A2L. **p < 0.01 Changes in cellular morphology as recorded using a widefield fluorescent microscopy. Resveratrol-induced morphological changes in HepG2 cells after 24 h of treatment (×20) demonstrate the significant time-and concentrationdependent antiproliferative capability of the ethanol leaf extract of A. leiocarpus (A2L). However, both resveratrol and the extract are non-cytotoxic except at high concentrations in which A2L is significantly more cytotoxic on HepG2 liver cancer cells than resveratrol in a time-dependent manner. Mechanisms of cell death by both treatments seem not to be ATPor antioxidant-depletion dependent.
[fig] Figure 5: Determination of positive nuclei total area. Changes in positive nuclear size associated with treatment of HepG2 cells with resveratrol or A2L followed up for three days [/fig]
[fig] Figure 6: Assessment of mitochondrial toxicity. Changes in the levels of cytotoxicity/membrane integrity (MI) and ATP determine if a treatment is mitotoxic [/fig]
|
Different Types of Errors in Saccadic Task Are Sensitive to Either Time of Day or Chronic Sleep Restriction
Circadian rhythms and restricted sleep length affect cognitive functions and, consequently, the performance of day to day activities. To date, no more than a few studies have explored the consequences of these factors on oculomotor behaviour. We have implemented a spatial cuing paradigm in an eye tracking experiment conducted four times of the day after one week of rested wakefulness and after one week of chronic partial sleep restriction. Our aim was to verify whether these conditions affect the number of a variety of saccadic task errors. Interestingly, we found that failures in response selection, i.e. premature responses and direction errors, were prone to time of day variations, whereas failures in response execution, i.e. omissions and commissions, were considerably affected by sleep deprivation. The former can be linked to the cue facilitation mechanism, while the latter to wake state instability and the diminished ability of top-down inhibition. Together, these results may be interpreted in terms of distinctive sensitivity of orienting and alerting systems to fatigue. Saccadic eye movements proved to be a novel and effective measure with which to study the susceptibility of attentional systems to time factors, thus, this approach is recommended for future research.
# Introduction
Throughout the day, cognitive performance stays under the combined influence of homeostatic sleep pressure and circadian processes. Interplay between these two factors affects the performance of everyday tasks and interferes with a contemporary lifestyle characterized by a large accumulation of duties which require continuous and prolonged cognitive efficiency. Such demands result in a reduction in sleep time, which has become a hallmark of modern society. Yet, most laboratory experiments deal with total sleep deprivation assuming that the consequences of chronic partial sleep restriction are comparable to those of total sleep loss [bib_ref] The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and..., Van Dongen [/bib_ref]. The effects of sleep restriction are a general decrease in alertness, wake state instability (including lapses of attention and microsleeps), perseveration of ineffective solutions, and weakened memory/learning performance [bib_ref] Neurocognitive consequences of sleep deprivation, Durmer [/bib_ref]. However, sustained attention deteriorates much more than the performance of challenging working memory tasks [bib_ref] Effects of partial and acute total sleep deprivation on performance across cognitive..., Lo [/bib_ref].
Even in a state of rested wakefulness, circadian rhythm driven by endogenous (eg. the circadian process and sleep pressure resulting in sleepiness and fatigue) as well as exogenous factors (eg. lighting and social environment) lead to variations in cognitive functioning [bib_ref] A time to think: circadian rhythms in human cognition, Schmidt [/bib_ref]. However, assessing human cognitive performance rhythms is complicated by two masking factors: the kind of task used and inter-individual differences in task performance [bib_ref] Circadian rhythms in cognitive performance: methodological constraints, protocols, theoretical underpinnings, Blatter [/bib_ref]. Moreover, the multidimensional nature of attention itself is probably the source of inconsistencies between the various studies of the circadian and homeostatic effects on performance relevant to the attentional domain investigated [bib_ref] A time to think: circadian rhythms in human cognition, Schmidt [/bib_ref] Errors arising due to sleep pressure and/or circadian rhythm may be related to either response selection or execution process. Their characteristics would vary depending on the neuronal network involved: alerting, orienting or executive control [bib_ref] The attention system of the human brain: 20 years after, Petersen [/bib_ref] , which are responsible for maintaining optimal vigilance, ordering sensory input and top-down regulation, respectively. Errors may also be differentiated depending on sleep pressure and the circadian process. As suggested by Valdez and colleagues [bib_ref] Circadian rhythms in components of attention, Valdez [/bib_ref] vigilance (sustained attention, concentration) is linked to fatigue, i.e. the homeostatic process, while tonic alertness (general capacity to respond), phasic alertness (the capacity to respond to a stimulus after a warning signal) and selective attention (the capacity to respond to a specific stimulus and ignore others) show circadian variations.
Research on sleep deficit in terms of alertness and vigilance has concentrated mostly on manual responses. The few experiments that have addressed oculomotor responses, have focused mostly on the velocity and latency of eye movements (eg. [bib_ref] Effect of 24-h sleep deprivation on the oculomotor reactions of a human..., Bukhtiyarov [/bib_ref]. The typology of the saccadic reactions is little explored-especially in the context of chronic sleep loss or circadian rhythms. However, saccadic responses provide a valid model for investigating basic cognitive processes, especially attentional capacities, and flexible control over behaviour. They are controlled by different low-and high-level systems [bib_ref] The three-loop model: a neural network for the generation of saccadic reaction..., Fischer [/bib_ref] and by separate excitatory and inhibitory neural pathways [bib_ref] Look away: the anti-saccade task and the voluntary control of eye movement, Munoz [/bib_ref]. These neural pathways are strictly linked with the top-down and bottom-up attentional processes, influencing eg. cue facilitation mechanisms in saccadic eye movement [bib_ref] Control of goal-directed and stimulus-driven attention in the brain, Corbetta [/bib_ref]. Thus, different types of saccadic task errors should be considered failures of diverse attentional functions.
The aim of this study was to verify the effects of chronic sleep deficit and time of day on human saccadic task performance. We hypothesized that those conditions would differentiate the oculomotor behaviour, and, in consequence, provide new information about the functioning of attentional sub-systems.
# Materials and methods
## Participants
Twenty four paid volunteers participated in this experiment (12 females, mean age 22.7 ± 1.6 years). Participants met the experiment requirements: right-handedness, right-eyed, normal or corrected-to-normal vision, no physical or psychiatric disorders. They were all non-smokers and drug-free. The absence of sleep problems or excessive sleepiness was confirmed by the Pittsburgh Sleep Quality Index [bib_ref] The Pittsburgh sleep quality index: A new instrument for psychiatric practice and..., Buysse [/bib_ref] , and the Epworth Sleepiness Scale [bib_ref] A new method for measuring daytime sleepiness: the Epworth sleepiness scale, Johns [/bib_ref]. Mean sleep quality index was 4.08 (std. err. 0.46) and the daytime sleepiness score was 5.71 (std. err. 0.53). Neither extremely morning-nor evening-oriented subjects were qualified for the study (Chronotype Questionnaire; [bib_ref] Can you feel the rhythm? A short questionnaire to describe two dimensions..., Oginska [/bib_ref].
# Ethics statement
Participants were informed about the procedure and goals of the study and gave their written consent. The study was approved by the Bioethics Commission at the Jagiellonian University.
## Procedure
The experimental task was performed in two conditions: rested wakefulness (RW-after a week with unrestricted, fully restorative sleep) and chronic sleep deficit (SD, after seven days of sleep curtailment). The order of experimental sessions (RW and SD) was counterbalanced across all participants. The sessions were separated by at least two weeks in order to minimize the residual effects of sleep deficit on performance. In the SD condition, the participants were instructed to shorten their sleep by delaying sleep time and using an alarm clock in the morning. The sleep length was individually calculated for them as two thirds of the self reported optimal sleep duration. The actual sleep duration and the sleep timing were verified on the basis of movement recorded by a Micro Motionlogger SleepWatch (Ambulatory Monitoring, Inc., Ardsley, NY) which was worn during the week before each experimental day on the participant's non-dominant wrist. According to actigraphy measurements, the sleep length in the SD condition was reduced by 30% in comparison to the RW condition. The severity of subjective sleep loss consequences was assessed with a CHICa scale [bib_ref] Measuring individual vulnerability to sleep loss -the CHICa scale, Oginska [/bib_ref].
All the participants performed an experimental task four times during the day: at approximately 10:00, 14:00, 18:00 and 22:00. Before each measurement session, participants were asked to estimate their alertness using the Karolinska Sleepiness Scale (KSS; [bib_ref] Subjective and objective sleepiness in the active individual, Akerstedt [/bib_ref]. A semi constant routine protocol was applied: room temperature and light intensity were kept constant, caloric intake and the level of motor activity were controlled. The participants spent approximately 14 hours in a controlled laboratory environment. During experimental days, they were allowed to engage in non-strenuous activities (eg. reading, watching videos, conversation). Caffeine intake was banned; alcohol consumption during the preceding week was restricted.
## Stimuli and task
A modified spatial cueing task [bib_ref] Orienting of attention, Posner [/bib_ref] was prepared using E-Prime 2.0 (Psychology Software Tools) and presented on a 17-inch screen located approximately 80cm from participants' eyes. Targets and cues were presented in six possible locations [fig_ref] Fig 1: Experimental task used in the study [/fig_ref] at 8°and 2°of visual angle in x-axis and 5°and 1°in y-axis of visual angle respectively. Leftwards and rightwards target locations were distributed equally, whereas middle target locations were weighted by 50% vs. 25% of upper and 25% of lower locations. The task comprised congruent trials with target stimuli preceded by congruent directional cues (60%, [fig_ref] Fig 1: Experimental task used in the study [/fig_ref] , incongruent trials with target stimuli preceded by incongruent directional cues (15%, [fig_ref] Fig 1: Experimental task used in the study [/fig_ref] , and stimuli without cues (25%, [fig_ref] Fig 1: Experimental task used in the study [/fig_ref]. The total number of stimuli in the task was 500 in each measurement. Targets were presented for 500ms and cues for 200ms. The intertrial interval was varied in the range of 80020133500ms with average of 2200ms. Time interval between cue and target varied between 200 and 700ms with an average of 450ms. The participants were instructed to direct their attention and gaze from fixation point to targets only if they were preceded by a cue. The task lasted about 35 minutes. One week prior to the first experimental day, participants were extensively trained on the experimental task, in order to avoid the influence of a learning process on the number of errors.
## Eye tracking data acquisition and analysis
Eye position was monitored with Smart Eye Pro (Smart Eye AB, Göteborg, Sweden). This eye tracking system uses the reflections of infrared flashes on the cornea to find the centre of eyes and calculate gaze direction. It has a 60 Hz sampling frequency, a headtracking range ±65°and eye tracking range ±20°of visual angle and a tracking accuracy of 0.5°. The analysis of eyetracking data was done using the Python language-based program, written for the purpose of the study. Saccades were detected using velocity and distance criteria, i.e. the movement was classified as a saccade only when its velocity reached 35°/s and its distance was at least 2°of visual angle. Oculomotor reaction time was calculated as a difference between target appearance and the beginning of a saccade.
Subjects were instructed to react only to targets preceded by a cue. Saccadic reactions were classified according to following criteria. The reactions to a target preceded by a cue (either congruent or incongruent) were recognized as correct (HIT) if the direction of the eye movement was consistent with the target position and if the reaction time was at least 80ms. Saccades executed earlier than 80ms after the stimulus onset (i.e. anticipatory saccades; [bib_ref] The three-loop model: a neural network for the generation of saccadic reaction..., Fischer [/bib_ref] and those before target appearance, regardless of the direction, were labelled premature reactions (PR). A lack of reaction during the stimulus presentation was considered as an omission error (OM). If the direction of the first saccade was opposite to the target location, the reaction was classified as a direction error (DirERR). In trials without a cue, the lack of any saccade was considered a correct rejection (CR). Reaction to the target not preceded by a cue was labelled as an error of commission (COM). Examples of reactions are depicted on [fig_ref] Fig 2: Representative reactions in saccadic task for rightward centered stimuli [/fig_ref] Recordings in which the eye tracking signal was poor and classification was not possible were discarded from further analyses. This resulted in an average loss of 6.4% (std.err. 1.3%) of trials.
# Results
The mean length of sleep during the RW condition (8h 9min ± 37min) was significantly longer (t (23) = 15.66, p < 0.001) than the average length of sleep during the SD condition (5h 41min ± 29min). On average, participants in the SD condition restricted their sleep time by Overall saccadic task performance measurements are depicted in [fig_ref] Table 1: Performance in saccadic task [/fig_ref]. The number of each type of reactions was analysed as a proportion of the overall number of classified reactions. An analysis of the variance (two-way ANOVA) with sleep condition and time of day (TOD) as factors was performed separately for each type of reaction. The analysis showed that the number of DirERRs changes significantly with TOD (F (3,69) = 3.51, p = 0.02; [fig_ref] Fig 3: Significant time of day variations for two types of reaction [/fig_ref]. A post-hoc LSD test showed that this kind of error was committed significantly more frequently in the first than in the second (p = 0.02) and the fourth session (p = 0.003). Considering that the abundance of DirERRs were committed on incongruent trials, additional analysis was conducted solely on these trials. The results presented similar TOD variations (F (3,69) = 3.32, p = 0.025) to those obtained for both types of trials. The proportion of PR in the number of all trials increased significantly depending on the TOD (F (3, 69) = 5.36, p = 0.002; [fig_ref] Fig 3: Significant time of day variations for two types of reaction [/fig_ref]. A posthoc test showed significant difference between the morning session and other times of day: 14:00 (p = 0.02), 18:00 (p = 0.002), and 22:00 (p = 0.0004).
The COM and OM showed no significant variation across different times of day. The ANOVA revealed a significant difference in the number of these errors between RW and SD conditions. The ratio of OM was significantly higher in the SD condition (F (1,23) = 13.61, p = 0.001; as was the proportion of COM (F (1,23) = 5.86, p = 0.02; . There was no significant interaction of TOD and sleep conditions for any type of reaction.
Changes in subjective sleepiness measured by KSS before each session showed significant variance between RW and SD condition [bib_ref] The cumulative cost of additional wakefulness: dose-response effects on neurobehavioral functions and..., Van Dongen [/bib_ref] = 68.62, p < 0.001) and between the times of the day (F (3,69) = 8.21, p < 0.001). There was no significant difference between average reaction times, neither between the two sleep conditions, nor between times of the day. Mean latency for all reaction are presented in [fig_ref] Table 1: Performance in saccadic task [/fig_ref].
# Discussion
Following a basic distinction of erroneous responses, errors can be caused both by failures in response execution, when the selected response program is correct but its implementation goes sideways, and failures in response selection, when a wrong, not consistent with the task demands, response program is executed [bib_ref] Detection of errors during speech production: a review of speech monitoring models, Postma [/bib_ref]. In this study, we were able to distinguish four types of errors lying beneath those categories. Failures in response execution took place when participants reacted to the cued target before its appearance (PR) or when they made a saccade to the opposite direction of the target location (DirERR). Accordingly, failures in response selection were observed either when participants shifted attention to uncued targets, while, according to the instruction, they were supposed to maintain focus on the fixation point (COM) or when Note: * indicate significant difference between reaction times in comparison to HIT in congruent trials (p < 0.001); C = congruent, I = incongruent.
participants did not execute a reaction on a cued target (OM). The different nature of the aforementioned errors suggests them to be associated with separate attentional networks [bib_ref] The attention system of the human brain, Posner [/bib_ref]. Namely, the unsuccessful implementation of the action indicates the area of response execution was scanned ineffectively and the reaction was either inadequate or executed to a stimulus not considered a target. It can also be explained in terms of the failure of the orienting networkthe mechanism responsible for shifting focus of attention on a potentially relevant area where the stimulus may occur [bib_ref] The attention system of the human brain: 20 years after, Petersen [/bib_ref]. Respectively, failures in response selection are considered as a result of disruption in alerting network activity [bib_ref] The effects of sleep deprivation on the attentional functions and vigilance, Roca [/bib_ref] , crucial to achieving and maintaining readiness to react to incoming stimuli. The results of our study showed the time of day variations in susceptibility to PR and Dir-ERR. These errors in response execution are related to the cue facilitation mechanism [bib_ref] Components of visual orienting, Posner [/bib_ref] as the cue leads to the orienting of attention and facilitates programming a saccade to the cued location. When the delay between the presentation of the gaze cue and the onset of a target is short, there is an increased tendency to commit DirERR caused by an automatic saccade following the incongruent cues. Facilitation leads not only to DirERR, but also to the anticipatory saccades, in our case the PR, linked with decision-related neural activity [bib_ref] Cortical control of ocular saccades in humans: a model for motricity, Pierrot-Deseilligny [/bib_ref]. The number of both DirERR and PR was not affected by chronic sleep loss, which is in line with the study of Martella et al. [bib_ref] Inhibition of return, but not facilitation, disappears under vigilance decrease due to..., Martella [/bib_ref] , showing that the facilitation effect is invulnerable to sleep deficit. Interestingly, the number of DirERR increases during the post-lunch dip [bib_ref] The post-lunch dip in performance, Monk [/bib_ref] and in the late evening, whereas the number of PR increases constantly during the day. The worsening of performance in the mid-afternoon has already been associated with the decreased efficiency of attention disengagement in the orienting process of attention shifts between different locations [bib_ref] Operation of attention disengagement and its diurnal variability, Fafrowicz [/bib_ref]. The worsening of performance in the late evening can be explained in terms of the fatigue-related failure in response execution caused by a crash of the orienting attention network. Thus, it can be concluded that the orienting attention sub-system is mostly prone to the circadian and fatigue factors which cause a boost in cue facilitation and prediction mechanisms and the increase of DirERR and PR. These findings are consistent with the results of our previous study [bib_ref] Diurnal patterns of activity of the orienting and executive attention neuronal networks..., Marek [/bib_ref] , which implemented neural measurements under the TOD condition and showed decreasing activity of the orienting attentional system.
The increase in the number of OM under the condition of chronic sleep deficit, similar to the acute condition, can be explained by the lapses of attention related to the wake state instability and microsleeps [bib_ref] Neurocognitive consequences of sleep deprivation, Durmer [/bib_ref] , during which the alerting sub-system of attention is compromised. This explanation is consistent with the evidence that at least subjective alertness and the ability to maintain sustained attention are highly influenced by sleep deprivation [bib_ref] Effects of partial and acute total sleep deprivation on performance across cognitive..., Lo [/bib_ref]. The increased number of COM, i.e. responses to no-go targets, is explained in terms of the diminished ability of the top-down inhibition of dominant automatic response [bib_ref] The Relations Among Inhibition and Interference Control Functions: A Latent-Variable Analysis, Friedman [/bib_ref]. This is consistent with the findings that response inhibition is particularly sensitive to sleep deprivation and occur even before an increment in the number of OM [bib_ref] Effects of two nights sleep deprivation and two nights recovery sleep on..., Drummond [/bib_ref]. It leads to inappropriate response selection regardless of other factors such as the circadian process and the age of sleep-deprived participants [bib_ref] Influence of age, circadian and homeostatic processes on inhibitory motor control: a..., Sagaspe [/bib_ref]. Taken together, the increase in the number of OM and COM is understood in term of changes in the interplay of two attentional processes: bottom-up and top-down, resulting in problems maintaining stable and adequate behavioural reactions [bib_ref] Sleep deprivation and vigilant attention, Lim [/bib_ref]. Considering that the errors in response selection can be explained in terms of the failure of the alerting sub-system of attention [bib_ref] The attention system of the human brain: 20 years after, Petersen [/bib_ref] , it can be concluded that this system is highly prone to chronic sleep restriction. It is in agreement with neuroimaging evidence, compensatory activation of the thalamus, the brain region linked with alertness, occurs under the condition of sleep deficit [bib_ref] Impairment of attentional networks after 1 night of sleep deprivation, Tomasi [/bib_ref].
The variations regarding both the alerting and orienting systems should be considered in relation to the supervisory system of executive control [bib_ref] The attention system of the human brain: 20 years after, Petersen [/bib_ref]. Indeed, a recent computational account [bib_ref] The impact of alertness on cognitive control, Nieuwenhuis [/bib_ref] indicated that complex behavioural patterns are a result of the interaction of these three attention systems. The impairment of the executive control system may lead to a failure of top-down processes and as a consequence, the disinhibition of bottom-up processes. On the one hand, the impairment of the executive control network influences the orienting one, as the former is linked with 'rapid strategic control over attention', enabling the voluntary switching of attention between cue and target locations [bib_ref] The attention system of the human brain: 20 years after, Petersen [/bib_ref] [bib_ref] Control of goal-directed and stimulus-driven attention in the brain, Corbetta [/bib_ref]. Thus, failures of the executive system may lead to changes in the functioning of orienting-related cue facilitation mechanisms, causing premature reactions and direction errors. On the other hand, the impairment of the executive control network also influences the alerting network, leading to increasing stimulus-driven reactions causing an increased number of commissions. Moreover, wake state instability and lapses of attention can be linked with the collapse of all attentional systems mirrored in the increased number of omissions [bib_ref] Neurocognitive consequences of sleep deprivation, Durmer [/bib_ref]. Taking our results into consideration, the functioning of the executive control network seems to be sensitive to both chronic sleep deficit and different times of day.
# Conclusions
Diverse attentional components and performance indices of sustained attention are differently affected by circadian and homeostatic processes (eg. [bib_ref] Circadian rhythms in components of attention, Valdez [/bib_ref] [bib_ref] Circadian and homeostatic variation in sustained attention, Valdez [/bib_ref]. Attention engagement and related performance are not based on a single cognitive process, but on various processes which can be differently affected by sleep restriction and circadian factors. In this study, we have shown that the time of day variations cause an increase in premature responses and direction errors, whereas chronic sleep deficit causes an increase in omissions and commissions. While the former can be related to the impairment of the orienting attentional system, the latter can be linked to failure of the alerting attentional network. Disruption in the functioning of these two systems can be caused by a decrement in the control ability mediated by the executive system. However, the above conclusions need further investigation, not only on a behavioural but foremost, on a neuronal level.
# Author contributions
[fig] Fig 1: Experimental task used in the study. (A) Possible locations for cues and targets; (B) Congruant trial; (C) Incongruent trial; (D) Trial without a cue. doi:10.1371/journal.pone.0126502.g001 Errors in Time of Day and Chronic Sleep Restriction PLOS ONE | DOI:10.1371/journal.pone.0126502 May 26, 2015 [/fig]
[fig] Fig 2: Representative reactions in saccadic task for rightward centered stimuli. Reaction types for (A) stimuli preceded by a cue and for (B) stimuli without a cue. HIT = correct reaction, DirERR = direction error, PR = premature reaction, OM = omission, CR = correct rejection, COM = commision. doi:10.1371/journal.pone.0126502.g002 Errors in Time of Day and Chronic Sleep Restriction PLOS ONE | DOI:10.1371/journal.pone.0126502 May 26, 2015 30% (± 8%) comparing to RW condition. The CHICa score differed significantly between the two conditions (RW: 9.0 ± 8.4, SD: 38.1 ± 16.0, t (23) = 8.62, p < 0.001). [/fig]
[fig] Fig 3: Significant time of day variations for two types of reaction: (A) DirERR; (B) PR. DirERR = direction error, PR = premature reaction. doi:10.1371/journal.pone.0126502.g003 [/fig]
[table] Table 1: Performance in saccadic task. [/table]
|
Development and Initial Validation of the PEG, a Three-item Scale Assessing Pain Intensity and Interference
BACKGROUND: Inadequate pain assessment is a barrier to appropriate pain management, but single-item "pain screening" provides limited information about chronic pain. Multidimensional pain measures such as the Brief Pain Inventory (BPI) are widely used in pain specialty and research settings, but are impractical for primary care. A brief and straightforward multidimensional pain measure could potentially improve initial assessment and followup of chronic pain in primary care.OBJECTIVES:To develop an ultra-brief pain measure derived from the BPI.DESIGN: Development of a shortened three-item pain measure and initial assessment of its reliability, validity, and responsiveness.PARTICIPANTS:We used data from 1) a longitudinal study of 500 primary care patients with chronic pain and 2) a cross-sectional study of 646 veterans recruited from ambulatory care.RESULTS: Selected items assess average pain intensity (P), interference with enjoyment of life (E), and interference with general activity (G). Reliability of the threeitem scale (PEG) was α=0.73 and 0.89 in the two study samples. Overall, construct validity of the PEG was good for various pain-specific measures (r=0.60-0.89 in Study 1 and r=0.77-0.95 in Study 2), and comparable to that of the BPI. The PEG was sensitive to change and differentiated between patients with and without pain improvement at 6 months.DISCUSSION:We provide strong initial evidence for reliability, construct validity, and responsiveness of the PEG among primary care and other ambulatory clinic patients. The PEG may be a practical and useful tool to improve assessment and monitoring of chronic pain in primary care.
# Background
Inadequate pain assessment has been identified as a key barrier to appropriate pain management. [bib_ref] Acute pain management: Operative or Medical Procedures and Trauma. Clinical Practice Guideline, Carr [/bib_ref] [bib_ref] Agency for Health Care Policy and Research, U.S. Department of Health and..., Jacox [/bib_ref] Recently, important initiatives have aimed to increase awareness of pain as a clinical problem by promoting better pain assessment. [bib_ref] Joint Commission pain standards: a progress report, Dahl [/bib_ref] These initiatives have led to widespread adoption of pain screening through measurement of current pain intensity.
In chronic pain, the most common type of pain seen in primary care, assessment of pain intensity alone is inadequate. Guidelines encourage comprehensive assessment that includes measurement of pain-related functioning, which may be even more relevant to patients' overall quality of life than intensity.To facilitate chronic pain assessment, numerous multidimensional patient-reported measures have been developed; 9-12 however, none of these have been widely adopted in the general medical settings where most chronic pain treatment is delivered.
In primary care, use of multidimensional pain measures is limited by factors such as instrument length and scoring complexity; however, a brief and straightforward multidimensional measure could potentially improve assessment of chronic pain. We sought to develop a very brief measure that would be feasible, valid, and sensitive to change in primary care. We started with the Brief Pain Inventory (BPI) because it is relatively easy to administer, score, and interpret; includes items assessing pain intensity and functional interference; and has been validated in many pain conditions. 10,13-17 As its name implies, the BPI is shorter than other multidimensional pain measures, but it is still too lengthy for implementation in primary care practice. We hypothesized that a shortened scale based on the BPI could be developed that would be more feasible, but just as useful, for assessing chronic pain in primary care. Our objectives were to develop an ultra-brief scale derived from the BPI and to initially assess its reliability, validity, and responsiveness.
# Methods
## Participants
We used data from two sources: 1) Stepped Care for Affective Disorders and Musculoskeletal Pain (SCAMP), a longitudinal study that enrolled a total of 500 patients with chronic musculoskeletal pain, and 2) Helping Veterans Experience Less Pain (HELP-vets), a cross-sectional study of 646 veterans receiving care at VA clinics. We used data from Study 1 to develop and initially validate the ultra-brief measure and data from Study 2 to confirm reliability and validity in an independent patient population. Study 1 (SCAMP) enrolled 500 primary care patients with persistent back, hip, or knee pain of at least moderate severity, 250 of whom had concurrent depression. [bib_ref] Stepped Care for Affective Disorders and Musculoskeletal Pain (SCAMP) study: design and..., Kroenke [/bib_ref] Participants were recruited from university (n=300) and VA-affiliated (n=200) internal medicine clinics in Indianapolis. Patients with concurrent depression were enrolled in a trial of depression and pain treatment vs. usual care (n=250). Those without depression were followed in a parallel observational study (n=250). The mean age of SCAMP participants was 59 years; 52% were women, 58% were white, and 38% were black. The mean numeric rating of current pain (on a 0-10 scale) was 6.1 (SD 1.9) at baseline. Study 2 (HELP-vets) enrolled a random visit-based sample of 646 veterans from ambulatory care clinics at two VA hospitals and six affiliated community sites in three urban California counties. Patients with chronic illness were over-sampled by design. The mean age was 63 years and 95% were male. Selfreported race/ethnicity was 54% white, 30% black, and 10% Latino. Sixty-one percent of participants reported pain at the time of enrollment and 63% had one or more pain diagnoses (33% back pain, 45% other musculoskeletal pain, 12% neuropathic pain, 5% headache). The mean rating of current pain (on a 0-10 scale) was 3.1 (SD 3.2) overall and 5.1 (SD 2.6) among those with pain.
## Measures
Study 1 participants completed the BPI, Chronic Pain Grade questionnaire (CPG), Roland disability scale, and SF-36 bodily pain scale at baseline; they completed the BPI, CPG, and pain global rating of change at 6 months. Study 2 was cross-sectional; participants completed the BPI, Functional Morbidity Index, and a single-item rating of overall pain-related distress.
& The Brief Pain Inventory (BPI) includes two scales that assess pain intensity and pain-related functional impairment (physical and emotional). [bib_ref] Dimensions of the impact of cancer pain in a four country sample:..., Cleeland [/bib_ref] The four items of the BPI severity scale assess the intensity of current pain and pain at its least, worst, and average during the past week on scales from 0 ("no pain") to 10 ("pain as bad as you can imagine"). The BPI interference scale assesses pain-related functional interference with seven items assessing different domains (general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life) rated from 0 ("does not interfere") to 10 ("interferes completely").
& The Chronic Pain Grade questionnaire (CPG) includes two three-item scales (intensity and disability) that are transformed into 0-100 scores. [bib_ref] Grading the severity of chronic pain, Korff [/bib_ref] An algorithm classifies pain into four graded categories: 1) low disability-low intensity, 2) low disability-high intensity, 3) high disability-moderately limiting, and 4) high disability-severely limiting. The CPG has been validated in primary care, chronic pain, and general populations. [bib_ref] Changes in chronic pain severity over time: the Chronic Pain Grade as..., Elliott [/bib_ref] [bib_ref] The Chronic Pain Grade questionnaire: validation and reliability in postal research, Smith [/bib_ref] [bib_ref] Back pain in primary care. Outcomes at 1 year, Korff [/bib_ref] & The Roland Disability questionnaire is a pain-specific measure of physical disability validated in patients with back pain and other chronic pain conditions. [bib_ref] A study of the natural history of back pain. Part I: development..., Roland [/bib_ref] [bib_ref] Validity of the Sickness Impact Profile Roland scale as a measure of..., Jensen [/bib_ref] It includes a checklist of 24 statements about pain effects on function; the score is the number of items endorsed.
& The Short-Form 36-item questionnaire (SF-36) Bodily Pain Scale is a two item scale assessing pain severity and interference. [bib_ref] The MOS 36-Item Short-Form Health Survey (SF-36): II. Psychometric and clinical tests..., Mchorney [/bib_ref] [bib_ref] The SF-36 Health Survey: development and use in mental health research and..., Ware [/bib_ref] Responses are transformed into a 0-100 score.
& The Pain Global Rating of Change is a single item assessing patients' overall impression of change in their pain. Study 1 participants were asked whether their pain was worse, about the same, or better since the start of the study. Those who reported that pain was better were asked to rate the magnitude of improvement (a little, somewhat, moderately, a lot, or completely better). Global ratings of change may be more sensitive to improvement and better correlated with patient satisfaction than serial measures. [bib_ref] Capturing the patient's view of change as a clinical outcome measure, Fischer [/bib_ref] & The Functional Morbidity Index was developed to assess general functional status in older adults. [bib_ref] Development and validation of a functional morbidity index to predict mortality in..., Carey [/bib_ref] Patients indicate whether they are able perform four different activities independently, and if not, whether the impairment is due to a health problem. & Overall Pain Distress is a single item: "How much did overall pain distress or bother you during the past week?" Response options are not at all, a little bit, somewhat, quite a bit, and very much.
## Item selection
We used a consensus-based process, drawing on a literature review, expert opinion, and statistical data, to develop a shortened scale. [bib_ref] Methodological approaches to shortening composite measurement scales, Coste [/bib_ref] Pre-specified criteria guided initial item selection. First, we decided to include at least one item representing each of three domains included in the BPI: pain intensity, physical functioning, and emotional functioning. We then selected items with the following characteristics: 1) easy to understand and applicable to patients with all types of pain; 2) good statistical characteristics (e.g., high response variability, high item-remainder correlation); 3) similar performance in depressed and non-depressed patients.
We chose "pain average" for the intensity item because it had a good distribution of responses, lacking the ceiling and floor effects seen with "pain worst" and "pain least," respectively. We did not select "pain now" because we wanted to avoid duplicating information provided by the "fifth vital sign," and capture intermittent pain. Although the ideal reporting period for pain assessment is debated, recalled average pain over one week is a valid measure of pain intensity. [bib_ref] Pain assessment in patients with low back pain: comparison of weekly recall..., Jamison [/bib_ref] BPI interference items include those assessing physical status (general activity, walking, normal work), emotional status (mood, relations with others, enjoyment of life), and sleep. For physical interference, we chose "interference with general activity" because it applies equally to all patients, as opposed to "interference with work" (which may be affected by occupation, employment status, etc.) and "interference with walking" (which may not apply to non-ambulatory patients or those with upper body pain).
For emotional interference, we considered both "interference with mood" and "interference with enjoyment of life." In our experience, "interference with relations with other people" is more difficult than other items for patients to answer. We wanted a scale that would discriminate between chronic pain and depression, which commonly co-occur. [bib_ref] Depression and pain comorbidity: a literature review, Bair [/bib_ref] In our sample of patients with and without comorbid depression, we found that "interference with enjoyment of life" was more independent of depression than "interference with mood." We also considered "interference with sleep" in place of the emotional interference items.
We reached consensus on a preferred three-item scale ("pain average," "interference with enjoyment of life," and "interference with general activity") and alternative three-item and four-item scales, which we then evaluated statistically.
## Reliability and validity
We assessed reliability (internal consistency) by calculating Cronbach's coefficient alpha. To assess construct validity, we compared the PEG with measures of pain and function using Pearson correlation coefficients. We used multiple measures for construct validity assessment, including the BPI, because no criterion standard exists for pain. We hypothesized that coefficients would be slightly higher for comparisons with pain-specific functional measures than for those with pain severity measures (because two of the three PEG items assess function). We also expected that coefficients would be larger for comparisons with pain-specific functional measures than for comparisons with generic functional measures.
## Responsiveness
Assessment of responsiveness, or sensitivity to change, requires an independent standard to define change. [bib_ref] Recommended methods for determining responsiveness and minimally important differences for patient-reported outcomes, Revicki [/bib_ref] We used two different measurements to define the presence or absence of patient improvement: 1) global rating of change and 2) serial CPG grade. We categorized patients according to their pain trajectory as assessed by each of the two measures. Global rating of change categories were defined by the patient's retrospective assessment at 6 months of the change in their pain since the trial began: 1) improved ("better"), 2) unchanged ("about the same"), and 3) worse ("worse"). CPG categories were defined by the change in CPG grade from baseline to 6 months: 1) improved (pain grade decreased by ≥1 level), 2) unchanged (pain grade at baseline = pain grade at follow-up), and 3) worse (pain grade increased by ≥1 level). Using data from Study 1, we assessed responsiveness by calculating the following three metrics: 1) change score (difference between mean score at baseline and follow-up), 2) effect size (ES; change score divided by the standard deviation of the baseline score), and 3) standardized response mean (SRM; change score divided by the standard deviation of the change score). These calculations were performed for patients in the improved, unchanged, and worse categories. Confidence intervals for SRM were calculated as + /-1.96 divided by the square root of the sample size. [bib_ref] Evaluating changes in health status: Reliability and responsiveness of five generic health..., Beaton [/bib_ref] We assessed responsiveness using all three methods because they can produce differing results and because agreement is lacking on the preferred method. [bib_ref] Evaluating changes in health status: Reliability and responsiveness of five generic health..., Beaton [/bib_ref] [bib_ref] A comparison of different indices of responsiveness, Wright [/bib_ref] We compared responsiveness of the PEG, BPI severity and BPI interference scales by comparing ES and SRM for each measure among patients in the improved category. Finally, we assessed responsiveness to varying degrees of improvement by comparing change scores for PEG and BPI scales to degree of improvement by global rating of change.
# Results
## Item selection
Using baseline data from the Study 1 sample, we assessed the preferred and alternate scales. Results were similar for all scales evaluated [fig_ref] Table 1: Reliability and Item-total Correlations for PEG and Alternate Scales in Study 1... [/fig_ref]. Our preferred scale demonstrated initial characteristics similar to or better than the alternatives, so we chose it as our final scale. From here on, we refer to this final abbreviated scale as the PEG, an acronym representing the three items: "Pain average," "interference with Enjoyment of life," and "interference with General activity." Principal components analysis of the PEG in both samples demonstrated a single factor, accounting for 66% of the variance in Study 1 and 81% in Study 2.
The PEG comprises 1 intensity item and 2 interference items [fig_ref] Figure 1: The PEG three-item scale [/fig_ref]. Consistent with BPI scoring, we calculated the average of individual item scores to get an overall PEG score (potential range 0-10). [fig_ref] Table 2: PEG and Individual Item Statistics at Baseline in Study 1 [/fig_ref] shows means and standard deviations (SD) for each item and the full three-item scale in both populations.
## Reliability and validity
Reliability of the PEG was 0.73 in Study 1 and 0.89 in Study 2. [fig_ref] Table 3: Correlation between the PEG, BPI Scales, and other Measures at Baseline [/fig_ref] shows correlation matrices for the PEG, BPI scales, and other pain and function measures in both study populations. Overall, construct validity of the PEG was good (r=0.60-0.89 in Study 1 and r=0.77-0.95 for pain-specific measures in Study 2), with correlations comparable to those of the BPI scales. As expected, PEG correlations were slightly higher for BPI interference (r=0.89 and 0.95) than for BPI severity (r=0.69 and 0.84) and for CPG disability (r=0.67) than for CPG intensity (r=0.64). Correlations were higher for the pain-specific function measures than for the Functional Morbidity Index (r =0.54), a generic measure of function.
## Responsiveness
Six-month follow-up data was available for 210 Study 1 clinical trial participants. The proportion with pain improvement was approximately the same according to global rating of change (31.4%) and serial CPG pain grade (29.5%); however, more patients were classified as worse by global rating (29.0%) than by CPG grade (15.0%). [fig_ref] Table 4: Responsiveness of PEG among Patients Classified by Pain Global Rating of Change... [/fig_ref] shows PEG scores and measures of responsiveness for patients classified as improved, unchanged, and worse at 6 months. Confidence intervals for the improved and unchanged groups did not overlap, but the unchanged and worse groups were not statistically different from each other. The improved group according to global rating of change had a mean improvement of 3.0 points (SD 2.5) on the PEG. Similarly, the
# Discussion
We demonstrated that the PEG, an ultra-brief three-item scale derived from the BPI, was a reliable and valid measure of pain among primary care patients with chronic musculoskeletal pain and diverse VA ambulatory patients. The PEG appears comparable to the BPI in terms of responsiveness to change. These findings support our hypothesis that an abbreviated scale derived from the BPI may be both useful and practical for chronic pain assessment in primary care and other ambulatory care settings, such as medical and surgical specialty clinics.
Strengths of this study include the confirmation of reliability and validity in an independent patient population, the diversity of the study populations, and the availability of multiple pain and functional measures with which to assess construct validity. Our choice of the BPI as the basis for our abbreviated scale development is another strength. The BPI is a widely used instrument that has been validated in numerous patient populations, clinical settings, and languages. BPI items are rated from 0-10, a format that has become familiar to patients and clinicians since assessment of pain with numeric scales has been broadly implemented in US health care settings. We took advantage of our collective experience with the BPI in observational and interventional research by employing a consensus-based process for scale shortening, consistent with recommendations to avoid over-reliance on statistical techniques. [bib_ref] Methodological approaches to shortening composite measurement scales, Coste [/bib_ref] We believe our use of two different ambulatory study populations is a strength; however, each study has its own limitations. Study 1 was a sample of patients with chronic back and lower extremity musculoskeletal pain and included an over-representation of patients with depression (50% by design). The more clinically diverse patient population of Study 2, including ambulatory VA patients with and without chronic pain, enhances the generalizability of our findings. However, Study 2 included fewer pain measures with which to assess construct validity and was cross-sectional; therefore, we were able to assess responsiveness only in the first sample. Forty percent of Study 1 patients and 100% of Study 2 patients were recruited from VA clinics, so our findings may be less generalizable to non-VA settings.
We found that the PEG differentiated well between patients who improved and those who did not. According to responsiveness metrics, patients in the improved category had a large improvement in PEG score, whereas those in the unchanged category had a minimal change. Proper interpretation of the magnitude of change according to SRM and one group pre-post ES is not entirely clear, although authors have suggested that Cohen's definition of small (0.2), moderate (0.5), and large (0.8) effects can be applied to interpretation of both responsiveness measures. [bib_ref] Comparative measurement sensitivity of short and longer health status instruments, Katz [/bib_ref] [bib_ref] Effect sizes for interpreting changes in health status, Kazis [/bib_ref] We did not find evidence of PEG responsiveness in the worse direction (i.e., change scores between those who were unchanged and those who were worse did not significantly differ). We are limited in our ability to adequately assess sensitivity to worsening because we evaluated responsiveness in a single study population that likely had a ceiling effect for worsening due to high baseline pain severity.
The competing demands of primary care, in which visits are short and pain is only one of several problems warranting attention, make efficiency of assessment a paramount concern. [bib_ref] Competing demands in psychosocial care. A model for the identification and treatment..., Klinkman [/bib_ref] [bib_ref] Is there time for management of patients with chronic diseases in primary..., Ostbye [/bib_ref] A balance must be found between feasibility and key characteristics such as reliability, validity, and responsiveness. For example, ultra-brief depression measures containing two to three items perform better than single item depression measures. [bib_ref] Do ultra-short screening instruments accurately detect depression in primary care? A pooled..., Mitchell [/bib_ref] We also evaluated a four-item abbreviated scale, but found that adding an item contributed little. An abbreviated version of the BPI that eliminated a few items has been previously published, [bib_ref] Reliability and validity of a modified Brief Pain Inventory short form in..., Mendoza [/bib_ref] but the PEG is the first ultra-brief scale based on the BPI.
New assessment strategies are needed to support improved chronic pain management in primary care. We believe the PEG, which includes items assessing pain intensity, emotional function, and physical function, is an important step forward. However, further studies are needed to confirm our findings and validate the PEG in additional patient populations. Prospective research should determine whether serial pain measurement can improve the quality of clinical decision-making and pain outcomes in primary care. Given the huge clinical and societal burden associated with pain, developing efficient and effective strategies to enhance care is an important priority.
[fig] Figure 1: The PEG three-item scale. *Items from the Brief Pain Inventory reproduced with permission from Dr. Charles Cleeland. [/fig]
[table] Table 1: Reliability and Item-total Correlations for PEG and Alternate Scales in Study 1 Sample (n=500) [/table]
[table] Table 2: PEG and Individual Item Statistics at Baseline in Study 1 [/table]
[table] Table 3: Correlation between the PEG, BPI Scales, and other Measures at Baseline [/table]
[table] Table 4: Responsiveness of PEG among Patients Classified by Pain Global Rating of Change and Serial CPG Grade at 6 Months (n=210) [/table]
|
Subsidising artemisinin‐based combination therapy in the private retail sector
## T a b l e o f c o n t e n t s
The ACT subsidy programmes increased the percentage of retail outlets stocking ACTs for children under five years of age by 31.9 percentage points (95% CI 26.3 to 37.5 percentage points; 1 study, high certainty evidence). E ects on ACT stocking for patients of any age is unknown because the certainty of evidence was very low.
The ACT subsidy programmes decreased the median cost of ACTs for children under five years of age by US$ 0.84 (median cost per ACT course without subsidy: US$ 1.08 versus with subsidy: US$ 0.24; 1 study, high certainty evidence).
The ACT subsidy programmes increased the market share of ACTs for children under five years of age by between 23.6 and 63.0 percentage points (1 study, high certainty evidence).
The ACT subsidy programmes decreased the use of older antimalarial drugs (such as amodiaquine and sulphadoxine-pyrimethamine) among children under five years of age by 10.4 percentage points (95% CI 3.9 to 16.9 percentage points; 1 study, high certainty evidence).
None of the three studies of ACT subsidies reported the number of patients treated who had confirmed malaria.
Vouchers increased the likelihood that an illness is treated with an ACT by 16 to 23 percentage points; however, vouchers were associated with a high rate of over-treatment of malaria (only 56% of patients taking ACTs from the drug shop tested positive for malaria under the 92% subsidy; 1 study, high certainty evidence).
# Authors' conclusions
Programmes that include substantive subsidies for private sector retailers combined with training of providers and social marketing improved use and availability of ACTs for children under five years of age with suspected malaria in research studies from three countries in East Africa. These programmes also reduced prices of ACTs, improved market share of ACTs and reduced the use of older antimalarial drugs among febrile children under five years of age. The research evaluates drug delivery but does not assess whether the patients had confirmed (parasite-diagnosed) malaria. None of the included studies assessed patient outcomes; it is therefore not known whether the e ects seen in the studies would translate to an impact on health.
## P l a i n l a n g u a g e s u m m a r y subsidising artemisinin-based combination therapy in drug shops and pharmacies
We conducted a review of the e ect of subsidising artemisinin-based combination therapy (ACT) drugs for malaria. We searched for all relevant studies up to February 2015 and identified four. Our findings are summarised below.
# Background
Malaria causes ill health and death in Africa, particularly in children under five years of age and poor rural populations. The World Health Organization recommends that people use ACT to treat malaria. ACT drugs are available at shops and pharmacies, but these drugs are expensive and people o en choose cheaper, older, less e ective drugs instead. The Global Fund and other international organisations have therefore decided to subsidise the cost of ACT drugs so that people can buy them from shops and pharmacies at prices similar to, or lower than, those of the older, less e ective drugs.
## What is the e ect of delivery programmes that subsidise act prices?
We included four studies. One study looked at the e ect of subsidising ACT drugs for children under five years of age and three studies looked at subsidising ACT drugs for people of all ages. All studies were from rural districts in East Africa (Kenya, Uganda and Tanzania). ACT price subsidies were accompanied with activities (such as sta training at shops and pharmacies, community awareness and mass Library Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews media campaigns) to promote appropriate use of antimalarial drugs in all except one study. In all four studies, the e ect of subsidising the drugs was compared to not subsidising the drugs. Price subsidies ranged from 80% to 95% of the actual price; vouchers to households were used in one study.
The findings from these studies indicate that ACT subsidy programmes:
(i) lead to a substantial increase in the number of children under five years of age who used ACTs when they had a fever (high certainty evidence);
(ii) lead to a substantial increase in the number of shops that stocked ACTs for children under five years of age (high certainty evidence); we could not draw any conclusion on the e ect on the number of shops that stocked ACTs for patients of any age because the quality of evidence was very low;
(iii) lead to a substantial decrease in the price of ACTs for children under five years of age (high certainty evidence);
(iv) lead to a substantial increase in the market share of ACTs for children under five years of age (high certainty evidence); and
(v) lead to a decrease in the use of older, less e ective antimalarials among children under five years of age (high certainty evidence).
None of the studies measured whether the subsidy programmes led to any harmful e ects (such as the inappropriate use of ACTs, in other words people who receive ACTs but do not actually have malaria).
The review findings also showed that subsidising ACT prices using vouchers lead to an increase in the likelihood that an illness was treated with an ACT among people seeking treatment for fever or suspected malaria. However, vouchers also lead to an increase in inappropriate use of ACTs (high certainty evidence).
## Illustrative comparative risks a (95% ci)
Assumed Cochrane Database of Systematic Reviews
## B a c k g r o u n d
Malaria is a major cause of ill health and death in Africa (WHO 2012). Uncomplicated Plasmodium falciparum malaria is the commonest form of the disease and accounts for most of the malaria cases and deaths . The World Health Organization (WHO) recommends artemisinin-based combination therapy (ACT) for the treatment of uncomplicated malaria (WHO 2012). These drugs are highly e ective and have the potential to reduce the development of antimalarial resistance. Unfortunately, despite the WHO's recommendation and substantial donor funding, only one in five antimalarial drugs used in malaria treatment in malariaendemic countries are ACTs . Reasons for the low use include high prices of ACTs in the retail sector (drug shops and pharmacies) where most people seek treatment for fever or suspected malaria (Patouillard 2010; Talisuna 2009). The price of ACTs is typically more than 10 times the price of older, less e ective antimalarial drugs such as amodiaquine (AQ) and sulphadoxinepyrimethamine (SP) in the retail sector (Morris 2014). The aim of the AMFm subsidy is to reduce ACT retail prices to a level similar to older, less e ective antimalarial drugs in order to increase demand and access for ACTs and displace artemisinin monotherapy and other sub-standard malaria treatments from the market, particularly among populations most vulnerable to malaria (children under five years of age and poor rural populations). Under the AMFm, 'first-line buyers' in the retail sector (those who buy ACTs directly from the manufacturer) pay about US$ 0.05 for a course of ACT rather than US$ 5.00 (the price paid before the AMFm; RBM 2007). The public sector can also purchase donor-subsidised ACTs, which may in turn broaden public sector ACT access. The subsidy programme is combined with supporting interventions (such as retail outlet provider training, community awareness and mass media campaigns) to facilitate e ective delivery and appropriate use of ACTs.
## Description of the intervention
## How the intervention might work
The AMFm subsidy programme is designed to increase access to ACTs in the retail and public sector by subsidising the prices of ACTs at the manufacturer level (Arrow 2004; Laxminarayan 2009). The programme aims to lower consumer prices of ACTs, compared to older and less e ective antimalarial drugs, available through the retail sector via two mechanisms: (1) negotiating with manufacturers of ACTs to reduce ACT prices; and (2) copaying a proportion (about 90%) of the reduced ACT price directly to participating manufacturers, hence further lowering prices to eligible wholesalers of ACTs (Global Fund; Laxminarayan 2009). The wholesalers thus pay a lower price for ACTs and prices fall all along the supply chain, increasing a ordability for the final consumer, while at the same time undercutting the price of resistanceinducing artemisinin monotherapy and competing with the prices for chloroquine and SP.
The pre-specified benchmarks of success of the AMFm pilot by the AMFm included: an increase in ACT use of 10 to 15 percentage points; increase in ACT availability of 20 percentage points; increase in ACT market share of 10 to 15 percentage points; and quality-assured ACT price less than 300% of the dominant non-quality-assured ACTs (chloroquine or SP; [fig_ref] Table 1: Guidelines for success benchmarks at 1 and 2 years a er e... [/fig_ref]. The AMFm process is illustrated in . In we provide a logic framework for this review, showing the theory of impact of the subsidies on malaria burden, potential influences on these steps and the outcomes that can be measured to evaluate the subsidy programmes.
## Library
Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews
## O b j e c t i v e s
To assess the e ect of programmes that include ACT price subsidies for private retailers on ACT use, availability, price and market share.
## M e t h o d s
Criteria for considering studies for this review
## Types of studies
Studies that assessed the e ect of retail sector ACT price subsidies using the following designs were eligible for inclusion (Appendix 1).
- Randomised trials.
- Non-randomised trials.
- Controlled before-a er studies.
- Interrupted-time-series studies (with a clearly defined point in time when the subsidy occurred, and at least three data points before and three a er the subsidy intervention).
## Types of participants
Studies involving the following groups of patients and units or channels of delivery of subsidised ACTs were eligible for inclusion.
- Consumers of retail sector subsidised ACTs (patients seeking treatment for fever or malaria; both children and adults). - Private retailers of subsidised ACTs (pharmacies, franchised clinics, drug shops, general stores).
## Types of interventions
## Intervention
Studies assessing the e ect of retail sector ACT price subsidies were eligible for inclusion.
- Retail sector ACT subsidy programmes (both for-profit and notfor-profit retail sectors). - Retail sector ACT subsidy programmes with supportive interventions (e.g. retail outlet provider training, community awareness and mass media campaigns).
## Comparisons
- Alternative ACT financing mechanisms aiming to achieve similar goals as retail sector ACT price subsidies (such as the generic Global Fund financing mechanism aiming to expand ACT availability in public health care facilities). - Public sector interventions to increase ACT availability funded by the United States President's Malaria Initiative (PMI) (http:// www.pmi.gov/). - Usual ACT delivery mechanisms (non-subsidised ACT interventions).
## Types of outcome measures
## Primary outcomes
ACT use (defined as the percentage of patients with fever/confirmed malaria who received an ACT on the day that the fever started or on the following day; Global Fund).
## Secondary outcomes
- ACT availability (proportion of all facilities stocking ACTs among outlets with any antimalarials in stock). - ACT price (cost to patients of a full child or adult course of ACTs).
- ACT market share (total volume of ACTs sold or distributed as a proportion of the total volume of all antimalarials sold or distributed via outlets). - Use of older antimalarial drugs (AQ, chloroquine, artemisinin monotherapy, SP). - Adverse e ects (such as the number of people receiving ACTs who do not have malaria).
All studies with eligible designs, participants and interventions were considered for inclusion irrespective of whether only the above outcome measures were reported.
## Search methods for identification of studies electronic searches
We aimed to identify eligible published and unpublished studies. We searched the following databases and grey literature sources in February, 2015.
-
## Additional resources
We also:
-
## Data collection and analysis
## Selection of studies
Two review authors (NO, PG) independently screened titles, abstracts and full texts of identified articles and applied the prespecified study eligibility criteria to select studies. GY reviewed all articles selected for inclusion. We resolved any disagreements by discussion. We documented the number of articles screened, assessed for eligibility and selected for inclusion in a PRISMA flow diagram Cochrane Database of Systematic Reviews
## Data extraction and management
Two review authors (NO, PG) independently extracted outcome data at baseline and endline. We also extracted data on study settings (coverage), participants (patient age groups), retail outlets, interventions (level of price subsidy and duration of subsidy programme), ACT supply (distribution) mechanisms and nature of supportive interventions (Characteristics of included studies; [fig_ref] Table 2: Characteristics of included studies Subsidising artemisinin-based combination therapy in the private retail... [/fig_ref]. Data were entered into a pilot-tested data extraction form. We resolved any disagreements by discussion.
## Assessment of risk of bias in included studies
Two review authors (NO, PG) independently assessed the risk of bias in the included studies using the Cochrane EPOC 'Risk of bias' tool (EPOC 2014). Quality domains assessed included: allocation sequence generation, allocation concealment, similarity of baseline characteristics and outcome measurements, blinding (personnel and outcome assessors), handling of incomplete outcome data, protection against contamination, completeness of follow-up and reporting of outcomes. We also assessed the following additional cluster-specific sources of bias: recruitment bias, loss of clusters, incorrect analysis and comparability with individually randomised trials (Higgins 2011). Findings were classified into three categories: low (low risk of bias for key quality domains, i.e. allocation sequence generation and concealment), high (high risk of bias for one or more of the key domains) and unclear (unclear risk of bias for one or more key domains). We resolved any disagreements by discussion.
We did not exclude studies on the basis of their risk of bias ratings; rather, we used these findings to help us better understand weaknesses in the identified evidence. We took into account the risk of bias ratings when synthesising and interpreting results. We report on the results of risk of bias assessment in the 'Risk of bias' tables and graphs ; .
## Library
Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews
## Figure 5. review authors' judgements about each risk of bias item presented as percentages across all included studies
## Measures of treatment e ect
We have reported measures of subsidy e ects as reported in the primary studies. We have presented absolute estimates of e ects (percentage point di erences, range and median e ect sizes) with 95% confidence intervals (CIs) where estimable. For example, for the outcome of ACT use, we reported percentage changes in control and intervention sites from baselines and the absolute percentage point di erence (cluster-adjusted) between the changes (i.e. 'di erence-in-di erence' estimates; [fig_ref] Table 3: Percentage of children with fever who received ACT on the same day... [/fig_ref].
## Unit of analysis issues
We assessed whether appropriate analysis was conducted to adjust for clustering in estimating precision of e ects in cluster trials (Higgins 2011).
## Dealing with missing data
We contacted authors of primary studies where relevant data were missing or where we required further clarification on the reported data. Where data were not available from the authors, we reported the data as missing; we did not impute or extrapolate values.
## Assessment of heterogeneity
We did not assess inconsistency between the results of individual studies using statistical methods; di erences in studies precluded meta-analysis. We documented factors that could modify subsidy e ects (such as programme coverage, level of price subsidy, ACT supply mechanism and nature of supportive interventions) in accordance with the established guidance for evaluating complex interventions (Shepperd 2009; [fig_ref] Table 2: Characteristics of included studies Subsidising artemisinin-based combination therapy in the private retail... [/fig_ref].
## Assessment of reporting biases
Reporting biases arise when the dissemination of research findings is influenced by the nature and direction of results (Sterne 2011). Reporting biases comprise publication bias, time lag bias, multiple (duplicate) publication bias, location bias, citation bias, language bias and outcome reporting bias. We assessed potential selective reporting of outcomes as one component of risk of bias assessment (we focused on the completeness of reporting of pre-specified outcomes). We did not create funnel plots as planned because of insu icient data (only four studies were included in the review) (Sterne 2011).
## Data synthesis
The included studies utilised varied study designs (randomised cluster and non-randomised cluster trials), enrolled diverse populations (children and adults) and used di erent e ect measures. We therefore did not combine results using statistical methods. We have described individual study results in the 'E ects of interventions' section.
## Library
Trusted evidence. Informed decisions. Better health.
## Cochrane database of systematic reviews
## 'summary of findings' table and assessing the certainty of evidence
We assessed the overall confidence in estimate of e ect (certainty of evidence) for each outcome using GRADE (Guyatt 2008). This system classifies the certainty of evidence (defined as 'the extent to which one can be confident that an estimate of e ect or association is correct') into four categories: very low, low, moderate or high. Data from randomised trials start at high quality while data from observational studies start at low quality. Quality of evidence from randomised trials can be downgraded in consideration of five factors: risk of bias or study limitations, directness, consistency of results, precision and publication bias. Similarly, quality of evidence from observational studies can be upgraded in consideration of three factors: magnitude of e ect estimate, doseresponse gradient and influence of residual plausible confounding.
Two review authors (NO, PG) independently assessed the certainty of evidence; we resolved disagreements by discussion. We did not assess the certainty of evidence for outcomes where there was insu icient data to permit reliable certainty rating. We did not exclude studies on the basis of GRADE ratings; we took into account the certainty of evidence when interpreting results.
We used GRADEpro so ware (GRADEpro 2015) to create 'Summary of findings' (SoF) tables for two comparisons: (1) 'retail sector ACT subsidies combined with supportive interventions versus no subsidies' (Summary of findings for the main comparison); and (2) 'ACT subsidy vouchers versus no subsidies' (Summary of findings 2). We included all the pre-specified outcomes in [fig_ref] Table 1: Guidelines for success benchmarks at 1 and 2 years a er e... [/fig_ref] and ACT access (defined as 'the share of illness episodes treated with ACTs') and targeting (defined as 'the share of ACT-takers who are malaria positive') in SoF table 2. We prioritised findings from randomised trials (data from non-randomised trials were incorporated as footnotes).
## Subgroup analysis and investigation of heterogeneity
We intended to perform subgroup analysis to investigate potential variation in subsidy e ects by study design, nature of supportive interventions, socioeconomic status and scale of coverage of subsidies (sub-national versus national programmes). However, we did not perform any of the planned analyses because there was insu icient data to permit reliable analysis.
# Sensitivity analysis
We intended to conduct sensitivity analysis to assess the impact of study quality on results (for example, whether e ect estimates are robust to changes in study risk of bias). We did not perform this analysis as no meta-analysis was done.
## R e s u l t s description of studies
## Results of the search
We identified a total of 1705 articles from both the electronic and supplementary searches. We excluded 1684 articles following a review of the titles and abstracts. We retrieved the full texts of 21 articles for detailed eligibility assessment. [fig_ref] Figure 3 ;: Moher 2009 [/fig_ref].
## Included studies
We included four studies (five articles; Characteristics of included studies; [fig_ref] Table 2: Characteristics of included studies Subsidising artemisinin-based combination therapy in the private retail... [/fig_ref] Two studies (three articles) were cluster-randomised trials ( ACT supply and distribution chains were varied: a third party procured and delivered ACTs directly to trained outlets from which shopkeepers purchased the treatment at a wholesale price in two trials . In one trial the project managers procured the ACTs directly from the manufacturing drug company (Novartis) and sold them to a wholesaler; drug shops purchased the ACTs from the wholesaler (Sabot 2009). Households were provided with vouchers to purchase subsidised ACTs from retail outlets in one trial (Cohen 2015).
## Randomised cluster trials
Kangwana 2011 evaluated the impact of retail sector ACT (artemether-lumefantrine; AL) subsidies in febrile children aged 3 to 59 months in Kenya. Nine areas were randomly allocated to the intervention arm (ACT subsidy plus supportive interventions, retail provider training and communication awareness activities) and nine to the control arm, with a bu er zone of two areas between selected areas. Subsidised AL was provided to retail outlets from November 2008. Cross-sectional household surveys were conducted before (July to August 2008) and a er (July to August 2009) the delivery of the intervention. The primary outcome was the proportion of children reporting fever in the previous two weeks who started treatment with AL on the same day as fever onset or the following day. Secondary outcomes were adequacy of AL doses obtained and consumed and the price paid per pack. Data were collected on 2749 children in the target age group at baseline and 2662 at one year follow-up.
Kangwana 2013 evaluated the impact of retail sector ACT subsidies using provider and mystery-shopper survey data collected as part of the randomised trial described above (Kangwana 2011). Data were collected at baseline (July to August 2008) and follow-up Library Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews (July to August 2009) using provider and mystery shopper crosssectional surveys. The mystery shopper survey assessed patientprovider interactions and aimed to provide data on actual rather than self-reported provider practice. Outcomes assessed included retail sector ACT availability, provider knowledge and provider dispensing practices. A total of 468 outlets were assessed at baseline and 639 at follow-up in the provider survey. 499 outlets were assessed at baseline and 653 at follow-up in the mystery shopper survey.
Cohen 2015 studied the impact of ACT vouchers in three rural districts in Kenya. Four drug shops (in four market centres) were selected and all households in the catchment area (within a 4 km radius) of these shops were sampled. Households were randomly assigned to one of three groups: (1) 'No subsidy' group (received vouchers to purchase full price ACTs at the pre-AMFm retail price of Kenya shillings 500 (approximately US$ 6.25 in 2009)); (2) 'ACT subsidy only' group (received vouchers for ACT subsidies of between 80% and 92%); (3) 'ACT plus rapid diagnostic test (RDT) subsidy' group (received vouchers for both subsidised ACTs and RDTs. Two vouchers for ACTs (AL) and two vouchers for RDTs (where applicable) were distributed to each household following a baseline survey; 2789 (95%) out of the 2928 households sampled consented to the baseline survey. The trial was conducted between May and December 2009 (the endline survey was administered about four months a er the vouchers had been distributed). We only extracted data on ACT access and targeting in the 'ACT subsidy' only and 'No subsidy' group. ACT access was defined as 'the share of illness episodes treated with ACTs'. ACT targeting was defined as 'the share of ACT-takers who are malaria positive'.
## Non-randomised cluster trials
Sabot 2009 assessed the impact of AMFm in three rural districts in Tanzania. Since two of the selected districts were adjacent, randomisation was limited so that one of the adjacent districts served as the control. The selected districts were randomly assigned to receive ACT subsidy, ACT subsidy plus suggested retail price or no ACT subsidy (control). The intervention was implemented in 2007. Data were collected at baseline and during intervention using interviews with drug shop customers, retail audits, mystery shoppers and audits of public and nongovernmental facilities. Most consumers interviewed in all districts were from the two least poor socioeconomic status quintiles (59% and 68%, respectively). A range of behaviour change communication (local radio advertisements and wall paintings) highlighting the importance of using ACTs and their availability in private shops was conducted by Population Services International.
Outcomes assessed were ACT uptake, availability and price. A total of 216 drug shops (duka la dawa baridi) were studied. The report focuses on data collected between August 2007 and August 2008.
Talisuna 2012 evaluated the impact of retail sector ACT subsidies in Uganda. The pilot was implemented in 2008 and involved four intervention districts (purposefully selected to receive branded subsidised ACTs) and one control district. Supportive interventions included communication and training activities to improve awareness of the availability of subsidised ACTs and correct dispensing and use of ACTs. Outcomes assessed were ACT uptake, availability, price, purchase within 24 hours of symptom onset and market share. Reported data comprise 1162 interviews at baseline and 5181 interviews at endline (May 2010) from 783 outlets.
## Excluded studies
We excluded 16 studies from the analysis because the study designs did not meet our inclusion criteria (uncontrolled pilots, pre-post surveys or qualitative assessments; Characteristics of excluded studies). Some of the studies excluded were large country evaluations without control sites where the observed e ects could be true e ects or could be secular and not due to the subsidy intervention (Tougher 2012; [fig_ref] Table 4: National ACT subsidy programmes a [/fig_ref]. Details of some of the excluded studies are described below.
Tougher 2012 assessed the e ect of AMFm in seven countries (Ghana, Kenya, Madagascar, Niger, Uganda, Nigeria and Tanzania (including Zanzibar)). Nationally representative baseline and endpoint surveys of public and private sector outlets that stock antimalarial treatments were conducted in each of the seven countries (eight national-level pilots). Clusters were selected using probability proportional to size sampling; independent samples were drawn at baseline and endpoint (a full census of outlets was done in Zanzibar because of the small population size). Outcomes assessed were ACT price, availability and market share. These outcomes were assessed against pre-specified success benchmarks a er one year of AMFm implementation [fig_ref] Table 1: Guidelines for success benchmarks at 1 and 2 years a er e... [/fig_ref]. Data on the implementation process and contextual factors (e.g. supportive interventions, mechanisms of distribution of co-paid ACTs) were collected through key informant interviews and document reviews.
Although this study provides important evidence on AMFm e ectiveness at a national scale we did not include it because it used a before-a er design with no comparator sites. The lack of control sites limits the degree to which observed e ects can be attributed to AMFm (findings may have been influenced by secular trends in measured outcomes and concurrent malaria interventions).
We identified six national programmes to scale-up subsidised ACTs in Cameroon, Senegal, Cambodia, DRC, Madagascar and Rwanda [fig_ref] Table 4: National ACT subsidy programmes a [/fig_ref]. Although the programmes report results indicative of the kind of e ect ACT subsidies can have under 'real world' conditions, we did not include them because they lacked comparison groups. In addition baseline data were not available for five of the national programmes (such data were only available for Rwanda's national programme).
## Risk of bias in included studies
We have presented findings on risk of bias assessment using 'Risk of bias' tables and graphs (Characteristics of included studies; ; . The overall risk of bias was low in two trials (
## E ects of interventions
See: Summary of findings for the main comparison ACT subsidies combined with supportive interventions versus no subsidies; Summary of findings 2 ACT price vouchers versus no subsidies
We have presented e ect estimates and certainty of evidence for each outcome in GRADE tables (Summary of findings for the main comparison; Summary of findings 2; Appendix 3; Appendix 4). We did not report the certainty of evidence for outcomes where there was insu icient data for GRADE assessment.
## Comparison 1: act subsidies combined with supportive interventions versus no subsidies
Three studies were included in this comparison .
## Primary outcome
Two studies reported data on ACT use (Kangwana 2011; Talisuna 2012; [fig_ref] Table 3: Percentage of children with fever who received ACT on the same day... [/fig_ref]. In the first study (Kangwana 2011), ACT subsidy programmes increased ACT usage in children under five years of age by 25 percentage points (95% CI 14.1 to 35.9 percentage points; high certainty evidence). This suggests that in practice, among febrile children under five years of age with an ACT usage rate of 5% without a subsidy, subsidy programmes would increase usage by between 19% and 41%. In the second study (Talisuna 2012), ACT subsidy programmes resulted in a ten-fold increase (95% CI 5.0 to 18.9) in ACT usage in children under five years of age and a six-fold increase (95% CI 4.2 to 8.4) in ACT usage in all age groups (very low certainty evidence).
## Secondary outcomes
Two studies reported data on ACT availability None of the three studies measured adverse e ects of ACT subsidy programmes.
## Comparison 2: act vouchers versus no subsidies
One study assessed the e ect of ACT vouchers to households on ACT access and targeting, among other outcomes (Cohen 2015). Compared to an access rate of 19% in the control group, subsidies of 80% or more increased the likelihood that a malaria-like illness was treated with an ACT by 16 to 23 percentage points (representing an 85% to 118% increase). However, subsidies were associated with a high rate of over-treatment of malaria: only 56% of patients taking ACTs from the drug shops tested positive for malaria under the 92% subsidy. The two lower subsidy levels were associated with much higher malaria positivity rates: "drug shop ACT-takers were 18 to 19 percentage points more likely to be malaria-positive under the 88 and 80 percent subsidies than under the 92 percent subsidy" (high certainty evidence).
## D i s c u s s i o n
## Summary of main results
This systematic review examined evidence from studies that evaluated the e ect of subsidy programmes aimed at improving accessibility and use of ACT for treatment of malaria. Four trials (five publications) that included substantive subsidies for private retailers were included in the review. All the studies were carried out in three adjacent countries in East Africa (Kenya, Uganda and Tanzania), and had accompanying interventions, including retail outlet provider training, community awareness and mass media campaigns.
The findings indicate that programmes that include substantive price subsidies (90% or more) for private antimalarial drug retailers improve use of ACTs (by 25 percentage points) among children under five years of age with suspected malaria. In practice this suggests that, among febrile children with an ACT usage rate of 5% without a subsidy, subsidy programmes would increase usage by between 19% and 41%. The findings also indicate that subsidy programmes improve ACT stocking in retail outlets (by 32 percentage points) and lower ACT prices (by US$ 0.84 per dose) for children under five years of age with suspected malaria. The impact on ACT stocking for patients of any age is unknown because the certainty of evidence was very low. Subsidy programmes also improve ACT market share (by between 24 to 63 percentage points) among patients of any age and reduce the use of older antimalarials (by 10 percentage points) among children under five years of age.
The findings also show that retail-sector ACT subsidies using vouchers lead to substantial increases in ACT access among people seeking treatment for suspected malaria. However, these subsidies also increase inappropriate use of ACTs (that is, increase the proportion of people who receive ACTs but do not in fact have malaria).
## Overall completeness and applicability of evidence
The data that were used in this review are from studies where malaria was mostly diagnosed based on the presence of fever in people seeking treatment for suspected malaria. The number of patients with confirmed (parasite-diagnosed) malaria was unclear in most of the studies; we therefore do not know whether subsidies resulted in over-treatment of malaria, except in the study of ACT vouchers where this was measured. None of the identified studies assessed mortality or other clinical outcomes. Furthermore, the studies did not examine patient adherence to subsidised ACTs. Thus, whether the observed improvement in ACT usage would translate into real health benefits remains uncertain. It is also unknown whether the impact of subsidies would vary by scale of coverage (sub-national vs. national subsidy programmes), ACT supply (distribution) mechanisms to retail outlets or socioeconomic status (there were insu icient data to explore e ects across these subgroups).
All the included studies were conducted in rural communities in low-income countries where malaria remains prevalent. Review findings are mostly generalisable to similar settings. However, the small scale nature of the subsidy programmes included in this review, di erences in the ACT supply mechanisms and retail sector distribution chains across settings may limit generalisability in some areas. Zambia, for example, has a more expansive public sector distribution chain compared to the private sector (Patouillard 2010).
Decisions to incorporate retail sector ACT subsidies into national malaria control programmes need to involve consideration of individual country contexts (which could include local malaria epidemiology, public sector to private sector antimalarial market ratio, diagnostic and distributional capacity of retail outlets, access to rapid malaria diagnostics and treatment seeking behaviours).
In addition malaria subsidy policies need to balance the benefits of retail sector ACT subsidies and potential unintended adverse e ects (for example, delaying the formal treatment-seeking that is needed for correct diagnosis and treatment of malaria and non-malarial fevers; under-treatment of malaria (under-dosing); failing to diagnose and treat co-morbid non-malarial fevers such as pneumonia; and over-treatment of malaria resulting from inappropriate use of subsidised ACTs in individuals with nonmalarial fevers). Such use of ACTs may increase the likelihood of emergence of artemisinin resistance.
## Quality of the evidence
The randomised trials provided high quality evidence on subsidy e ects . The certainty of evidence for all outcomes reported in the three trials was judged to be high (i.e. the research provides a very good indication of the likely e ect).
The findings from the other two studies were susceptible to bias . We downgraded the certainty of evidence (for ACT stocking for patients of any age) in one study because of high likelihood of selection bias (non-randomised evaluation) and possible confounding by study site (only one control site included; results were likely to be influenced by sitespecific factors; Sabot 2009).
We excluded several studies because they used ineligible designs prone to bias. The ideal designs to assess the e ects of large scale public health interventions such as ACT subsidy programmes are cluster-randomised trials with comparable control sites. However, such experimental designs are rarely feasible in practice. For example, cluster randomisation of regions in the included studies was limited by the need to use existing pharmaceutical retail distribution channels. Consequently, implementation of the subsidy interventions could not be restricted to certain areas as predicted by randomisation processes. Furthermore, identification of comparable control groups remains a challenge given inherent di erences in contexts such as health systems arrangements in malaria-endemic settings. This challenge was addressed in one included study (Kangwana 2013)
## Potential biases in the review process
We excluded many potentially eligible studies because of ineligible study designs. It is possible that some of these studies provide useful information that might complement findings from the four included studies. For example, positive e ects observed in the included studies were replicated in one excluded study (Tougher 2012). Consistent findings from di erent study designs across varying malaria transmission and cultural contexts increase our confidence that observed improvements in ACT use, availability, prices and market share can be attributed to the studied subsidy programmes.
We intended to include only randomised trials, non-randomised trials, controlled before-a er studies with at least two intervention and two control sites and interrupted-time-series studies (Appendix 1). These criteria were necessary to minimise possible confounding of subsidy e ects by site-specific factors (such as ACT supply chains, regulatory policies and retail provider behaviours). However, we made a post-hoc decision to include two nonrandomised cluster trials (Sabot 2009; Talisuna 2012) which compared intervention sites to only one control site. We therefore cannot rule out the possible influence of site-specific factors on observed subsidy e ects (hence the decision to downgrade certainty of evidence in Sabot 2009). In addition, government regulatory interventions to phase out monotherapy (AQ/SP) may have contributed to the observed decline in the use of these drugs in Kenya and Uganda (Kangwana 2011; Talisuna 2012). We also did not assess e ects on two outcomes as planned in Cochrane Database of Systematic Reviews our protocol: 'availability of alternative antimalarial drugs in all facilities, private and public (including informal outlets)' and 'prices of alternative antimalarial drugs (full adult or child courses)'; we used six outcomes included in the 'Summary of findings' tables.
## Agreements and disagreements with other studies or reviews
We identified one related review and one study of the impact of ACT subsidy programmes (Morris 2014; Tougher 2012).
Morris 2014 assessed the impact of retail sector ACT subsidies on ACT use. The review included 40 studies, comprising 10 experimental subsidies in eight countries, non-AMFm programmatic subsidies in nine countries and AMFm subsidies in eight pilots. Reported findings were derived from four experimental subsidies, three programmatic subsidies and five of the eight AMFm pilot subsidies. ACT subsidies substantially increased use of ACTs among patients with suspected malaria: each US$1 decrease in price was linked to a 24 percentage point increase in the fraction of suspected malaria cases purchasing ACTs. There were no di erences in ACT use among the poorest and richest groups, rural versus urban populations or children versus adults. The authors concluded that ACT price reductions can increase ACT use for suspected malaria, even within poorer, more remote populations that may be most at risk of malaria mortality.
Tougher 2012 assessed the e ect of the AMFm in seven countries (Ghana, Kenya, Madagascar, Niger, Uganda, Nigeria and Tanzania (including Zanzibar)). The study used a before-a er design with no comparator sites (see Excluded studies for details on study methods). The AMFm resulted in a large increase in quality-assured ACT (QAACT) availability (by 25.8 to 51.9 percentage points) in all pilots except Niger and Madagascar, and a large increase in ACT market share (by 15.9 to 40.3 percentage points), driven mainly by changes in the private for-profit sector. Median price for QAACTs per adult equivalent dose decreased substantially in the private for-profit sector in six pilots; the decrease ranged from US$1.28 to $4.82. The market share of oral artemisinin monotherapies decreased in Nigeria and Zanzibar, the two pilots where it was more than 5% at baseline. The authors concluded that subsidies combined with supporting interventions can be e ective in rapidly improving availability, price and market share of QAACTs, particularly in the private for-profit sector.
We did not conduct a quality assessment (risk of bias) for the evidence presented in the related review and study; these findings should therefore be interpreted with caution. However, despite di erences in study designs, the conclusions in both studies are consistent with the findings of our review: ACT subsidies combined with supportive interventions increase ACT usage, availability and market share and lower ACT prices for people seeking treatment for suspected malaria.
## A u t h o r s ' c o n c l u s i o n s
## Implications for practice
The findings of this review suggest that programmes that include substantive price subsidies (90% or more) for ACTs for private retailers combined with provider training and marketing improve use and availability of ACTs and lower ACT prices for children under five years of age with suspected malaria. This research has also shown that subsidy programmes improve market share of ACTs (volume of ACTs distributed as a proportion of total volume of all antimalarials distributed via outlets) and reduce use of older, less e ective antimalarials for children. We could not draw any conclusion on the impact on ACT stocking for patients of any age because the certainty of evidence was very low.
Decisions to incorporate retail sector ACT subsidies into national malaria control programmes need to involve consideration of individual country contexts and weigh the benefits of subsidies against potential unintended consequences (such as overtreatment of malaria resulting from inappropriate use of ACTs among patients with non-malarial fevers). E orts to scale-up retail sector ACT subsidy programmes in malaria-endemic settings (for example, via licensed community based pharmacies) should be complemented with policies to strengthen health systems (for example, enhanced malaria diagnostics using subsidised rapid diagnostic tests to improve ACT targeting; improved antimalarial drug supply in the public sector; in-service malaria case management training; and routine monitoring and surveillance for safety and impact).
## Implications for research
The number of patients with confirmed (blood-diagnosed) malaria was unclear in most of the included studies. Thus, future studies should investigate options to better target subsidised ACTs to patients who actually have malaria (for example, e ectiveness of retail sector ACT subsidies combined with subsidies for rapid diagnostic tests). Optimal targeting of subsidised ACTs would increase the likelihood that non-malarial illness such as pneumonia (the symptoms of which are o en similar to those of malaria) are promptly diagnosed, treated or referred. Such targeting would also reduce the likelihood of emergence of artemisinin resistance. These studies should ideally use cluster-randomised, interrupted-timeseries or plausibility designs (Victora 2004).
Future studies should also investigate pharmacovigilance and the extent of under-and over-treatment of malaria resulting from inappropriate targeting of retail sector ACT subsidies. The coste ectiveness and sustainability of subsidy programmes compared to alternative financing mechanisms and other approaches to expand access to subsidised malaria drugs (such as private retail sector-public partnerships and community based strategies) also need to be investigated.
## A c k n o w l e d g e m e n t s
We would like to thank Marit Johansen and John Eyers for their assistance with the literature searches and Andy Oxman for comments on the review manuscript. This review is supported by a project funded by UKAid from the UK Government Department for International Development.
## Mentor 2010 {unpublished data only}
## Who 2009
World Health Organization. World malaria report 2009. http:// www.who.int/malaria/world_malaria_report_2009/en/ (accessed 12 June 2013).
## Who 2012
World Health Organization. World Malaria Report 2012. http://www.who.int/malaria/publications/ world_malaria_report_2012/en/ (accessed 12 June 2013).
## World bank 2014
World Bank. Country classification. http://data.worldbank.org/ about/country-and-lending-groups (accessed 10 September 2014).
## C h a r a c t e r i s t i c s o f s t u d i e s
## Characteristics of included studies [ordered by year of study]
Methods Non-randomised controlled cluster trial The project managers procured AL from the manufacturer, Novartis, and sold them to a pharmaceutical wholesaler in Dar es Salaam at an average of US$ $0.11 per dose, 88% below the price offered to public buyers.
In one of the intervention districts (Kongwa), the suggested retail price intended to inform consumers of the maximum amount they should pay was set at 300, 600, 900, and 1200 Tanzanian shillings (0.25, 0.50. 0.75. and 1 USD respectively) for the four weight packs respectively; no suggested retail price was included on drugs distributed to Maswa in order to test its effect on price outcomes.
## Risk of bias
## Bias authors' judgement support for judgement
Random sequence generation (selection bias)
High risk Quote: "The selected districts were randomly assigned to one of the three arms in the study design: subsidy, subsidy plus suggested retail price, and no subsidy (control). As two of the qualified districts were adjacent, randomization was limited so that one of the adjacent districts served as the control"
Comment: Non-randomised design Allocation concealment (selection bias)
High risk Quote: "The selected districts were randomly assigned to one of the three arms in the study design: subsidy, subsidy plus suggested retail price, and no subsidy (control). As two of the qualified districts were adjacent, randomization was limited so that one of the adjacent districts served as the control"
## Sabot 2009
Subsidising artemisinin-based combination therapy in the private retail sector ( Blinding of participants and personnel (performance bias) All outcomes Low risk Comment: Blinding was not possible for study personnel and ACT providers due to the public information campaign around the subsidised drugs in the intervention arm. Lack of blinding was, however, unlikely to influence results as the outcomes assessed were objective.
Blinding of outcome assessment (detection bias) All outcomes Low risk Comment: Blinding was not possible for data collectors due to the public information campaign around the subsidised drugs in the intervention arm. Lack of blinding was however unlikely to influence results as outcomes assessed were objective.
## Contamination
High risk Quote: "As two of the qualified districts were adjacent, randomization was limited so that one of the adjacent districts served as the control"
Incomplete outcome data (attrition bias) All outcomes
Low risk Comment: The number of drug shops closed or refusing to participate were 30 (13%) and 39 (15%) respectively.
Selective reporting (reporting bias)
Low risk Comment: Data on pre-specified outcomes reported Recruitment bias High risk Quote: "The selected districts were randomly assigned to one of the three arms in the study design: subsidy, subsidy plus suggested retail price, and no subsidy (control). As two of the qualified districts were adjacent, randomization was limited so that one of the adjacent districts served as the control"
Loss of clusters Low risk Quote: "The total number of DLDB audited increased from 200 in August 2007 to 216 in August 2008 due to the opening of new shops, with 30 (13%) and 39 (15%) additional shops closed or refusing to participate at these two time periods respectively."
Incorrect analysis Low risk Quote: "To assess geographical variation in outcomes, the competition level of all DLDB was calculated using the fixed radius approach...The competitive space of each DLDB was defined as 1 kilometer and each shop was assigned to a competition index category between 0 and 5 based on the number of other DLDB within that radius." "A repeated measures multivariate regression model was used to compare differences in purchase price while controlling for potentially confounding factors and adjusting for clustering of multiple purchases in the same shops."
Comparability with individually randomised trials Supportive interventions: Training of retail outlet sta , job aids, community awareness activities (e.g. workshops, posters and paintings on shops; these activities were designed to make the community aware of malaria, the availability of Tibamal, and the importance of adherence to the medication).
## Outcomes
Primary outcome: proportion of children reporting fever in the past 2 weeks who started treatment with AL on the day or following day of fever onset. Secondary outcomes: adequacy of AL doses obtained and consumed and the price paid per pack
## Notes
At the time of the study, AL had a retail price of around US$ 6.16 (500 Kenyan shillings) compared with an average of around US$ 0.37 for common, older antimalarials such as SP and AQ. The outlets were instructed to sell the packs at a retail price of US$ 0.25, which was printed on the drug packaging, providing a 150% retailer mark-up (exceeding that of AQ and SP, which generally had retail markups of 50% to 100%).
## Risk of bias
## Bias authors' judgement support for judgement
Random sequence generation (selection bias) Low risk Comment: A random list of all eligible sublocations was formulated per district in Microsoft Excel. The first intervention sublocation was selected from the top of the list. In order to reduce the potential for contamination a ''bu er zone'' was created where all sublocations located within two sublocation boundaries of the selected sublocation were removed from the list. The list was reshuffled randomly and the first sublocation on the new list allocated to the control arm.
## Allocation concealment (selection bias)
Low risk Comment: The same procedure as for random sequence generation (described above) was used; intervention allocation could not have been seen in advance.
Baseline outcome measurements Low Supportive interventions: Communication activities to improve awareness of the importance and availability of ACTs, and training activities to ensure correct dispensing and use of subsidised ACTs Outcomes ACT uptake, purchase of ACT within 24 hours of symptom onset and ACT price, availability and market share Notes There was better availability of ACT in the public sector in the control district because of: (1) new interventions initiated targeting the community level distribution of ACTs through the public sector; and (2) improvements in the procurement and distribution system in the public sector, based on a push instead of a pull system for the lower level health units.
The maximum recommended retail price for the subsidised ACT ranged from 200 Ugandan shillings to 800 Ugandan shillings (US$ 0.10 to 0.40), depending on the target age/doses.
## Library
Trusted evidence. Informed decisions. Better health.
## Cochrane database of systematic reviews
## Bias authors' judgement support for judgement
Random sequence generation (selection bias) High risk Quote: "Four intervention districts were purposefully selected to receive branded subsidised medicines -'ACT with a leaf', while the fi h district acted as the control."
Comment: Non-randomised design Allocation concealment (selection bias)
High risk Quote: "Four intervention districts were purposefully selected to receive branded subsidised medicines -'ACT with a leaf', while the fi h district acted as the control."
Comment: Non-randomised design Baseline outcome measurements Low risk Comment: No important differences across study groups on pre-specified subsidy outcome measures Baseline characteristics Low risk Quote: "Some discordance was observed between the intervention and control districts at baseline in terms of drug consumption habits. Fortunately, the observed disparities did not include the use of ACT. Generating survey-adjusted outputs was intended to provide a more reasonable range of likely values that accounted for this prior to executing tests of significance"
Blinding of participants and personnel (performance bias) All outcomes Low risk Comment: Blinding was not possible for drug shop personnel due to the public information campaign around the subsidised drugs in the intervention arm. Lack of blinding was, however, unlikely to influence outcomes as outcomes assessed were objective.
Blinding of outcome assessment (detection bias) All outcomes Low risk Quote: "Possible interviewer bias minimised through a week long training to instil strict processes for conducting interviews, and to minimise deviation from the interview script. It was unlikely that the lack of blinding could influence outcomes as it was obligatory for the interviewer to observe and record the details of the actual medicine purchased."
## Contamination
Low risk Quote: "To limit leakage of the intervention to the control area, the control and intervention areas had two intervening bu er districts (Bukedea and Kumi) or a lake between them."
Incomplete outcome data (attrition bias) All outcomes Low risk Quote: "...the rate of refusal was generally small (not exceeding 10%) and any effect due to refusal probably did not impact significantly on the outcome measures." Comment: Of the 5,643 observations collected in the final evaluation survey, 5,181 observations resulting from visits to 783 outlets were included in the analysis.
Selective reporting (reporting bias)
Low risk Comment: Data on pre-specified outcomes reported Recruitment bias Low risk Quote: "Four intervention districts were purposefully selected to receive branded subsidised medicines -'ACT with a leaf', while the fi h district acted as the control." Cochrane Database of Systematic Reviews 5 interviews per outlet were respected in the final survey round, resulting in a significantly higher sample size."
## Loss of clusters
Incorrect analysis Low risk Quote: "Based on the data-collection methods, a survey-adjusted logistic regression model was used. The outlets were treated as the population sampling units within five strata -the five pilot districts."
Comparability with individually randomised trials Supportive interventions: Training of retail outlet sta , job aids, community awareness activities (e.g. workshops, posters and paintings on shops; these activities were designed to make the community aware of malaria, the availability of Tibamal, and the importance of adherence to the medication).
Outcomes AL uptake (provider behaviour), availability of older antimalarials, AL price, AL stocking, provider knowledge, and provider dispensing practices. Generally, outlets that received subsidised AL plus training and job aids performed better than those receiving one or none of these intervention components.
## Notes
At the time of the study, AL had a retail price of around US$ 6.16 (500 Kenyan shillings) compared with an average of around US$ 0.37 for common, older antimalarials such as SP and AQ. The outlets were instructed to sell the packs at a retail price of US$ 0.25, which was printed on the drug packaging, providing a 150% retailer mark-up (exceeding that of AQ and SP, which generally had retail markups of 50% to 100%). Generally, outlets that received training and job aids performed better than those receiving one or none of these intervention components.
## Risk of bias
## Bias authors' judgement support for judgement
Random sequence generation (selection bias) Low risk Comment: A random list of all eligible sublocations was formulated per district in Microsoft Excel. The first intervention sublocation was selected from the top of the list. In order to reduce the potential for contamination a ''bu er zone'' was created where all sublocations located within two sublocation boundaries of the selected sublocation were removed from the list. The list was reshuffled randomly and the first sublocation on the new list allocated to the control arm.
## Library
Trusted evidence. Informed decisions. Better health.
## Cochrane database of systematic reviews
Allocation concealment (selection bias)
Low risk Comment: The same procedure as for random sequence generation (described above) was used; intervention allocation could not have been seen in advance.
Baseline outcome measurements Low risk Comment:No important differences across study groups on pre-specified subsidy outcome measures Baseline characteristics Low risk Quote: "To control for potential confounders the covariates considered were outlet type (specialized drug store or general store), distance of shop to nearest road, whether any sta had clinically related training, and district. All covariates significant at a p-value of <0.2 were retained in the regression model."
Blinding of participants and personnel (performance bias) All outcomes Low risk Comment: Blinding was not possible for shopkeepers due to the public information campaign around the subsidied drugs in the intervention arm. Lack of blinding was, however, unlikely to influence results as the outcomes assessed were objective.
Blinding of outcome assessment (detection bias) All outcomes Low risk Comment: Blinding was not possible for data collectors due to the public information campaign around the subsidised drugs in the intervention arm. Lack of blinding was, however, unlikely to influence results as outcomes assessed were objective.
## Contamination
Low risk Quote: "It is also possible that there was some contamination of the control arm outlets, which could have heard some of the communication activities. However, results indicated that such exposure was low, with only 1% of control arm respondents saying that they had attended the Tibamal training, 14% having heard of Tibamal, and no outlets stocking Tibamal."
Incomplete outcome data (attrition bias) All outcomes Low risk Quote: "In both the provider and mystery-shopper surveys, at baseline and also at follow-up, less than 10% of outlets were not interviewed either because the respondent refused to be interviewed or the outlet was closed during visits."
Selective reporting (reporting bias)
Low Households were randomly assigned to one of three groups:
(1) "No subsidy" group ("received vouchers to purchase unsubsidized ACTs at the market price of KSh (Kenyan shillings) 500 (just under $6.25). This treatment arm was meant to capture the no-subsidy status quo that prevailed in Kenya prior to the AMFm pilot, in which over-the-counter ACTs were expensive and RDTs were not available in drug shops.")
(2) ACT subsidy only group
(3) 'ACT plus RDT' subsidy group (received vouchers for both subsidised ACTs and RDTs)
The ACT used in the study was Coartem (AL).
Supportive interventions: None.
## Outcomes
Outcomes of interest to current review:
ACT accessibility ("the share of illness episodes treated with ACTs");
ACT targeting ("the share of ACT-takers who are malaria positive")
Notes "Within the two ACT subsidy groups ('ACT subsidy only' and 'ACT+RDT subsidy'), households were randomly assigned to an ACT subsidy level of 92, 88 or 80 percent (corresponding to $0.50, $0.75 and $1.25 for an adult dose, respectively). The 92 percent subsidy level corresponds to the Kenyan government's target retail price of KSh 40 under the AMFm."
## Risk of bias
## Bias authors' judgement support for judgement
Random sequence generation (selection bias)
Low risk Quote: "The randomization of households was done using a computerized random number assignment algorithm and was stratified by drug shop, by the household's distance to the drug shop (in quartiles) and by the presence of children in the household."
Allocation concealment (selection bias)
Low risk Comment: The same procedure as for random sequence generation (described above) was used; selection bias due to foreknowledge of treatment allocation is considered unlikely.
Baseline outcome measurements Low risk Quote: "There are no significant differences across treatment groups, other than for the number of acres owned and the age distribution in the household. In particular, our control group has slightly older household heads, with, as a consequence, a significantly higher fraction of adults. Since age is highly correlated with malaria positivity, a lack of balance across treatment groups in the age composition of households could confound estimates of treatment assignment on uptake and targeting, even though the magnitude of the age differences is not large. Therefore, unless otherwise noted, we control for the age of the household head in all of our results. "
Baseline characteristics Low risk Quote: "There are no significant differences across treatment groups, other than for the number of acres owned and the age distribution in the household. In particular, our control group has slightly older household heads, with, as a consequence, a significantly higher fraction of adults. Since age is highly correlated with malaria positivity, a lack of balance across treatment groups in the age composition of households could confound estimates of treatment assign- Cochrane Database of Systematic Reviews ment on uptake and targeting, even though the magnitude of the age differences is not large. Therefore, unless otherwise noted, we control for the age of the household head in all of our results."
Blinding of participants and personnel (performance bias) All outcomes Low risk Comment: Lack of blinding considered unlikely to impact on outcomes
Blinding of outcome assessment (detection bias) All outcomes Low risk Comment: Lack of blinding considered unlikely to impact on outcomes Contamination Low risk Comment: Households were provided with vouchers for subsidised and nonsubsidised ACTs; cross-group contamination is considered unlikely.
Incomplete outcome data (attrition bias) All outcomes Low risk Quote: "Only five percent of households surveyed at baseline were not reached at endline, and attrition was balanced across treatment arms."
Selective reporting (reporting bias)
Low risk Comment: Data on pre-specified outcomes were reported.
## Recruitment bias low risk
Quote: "We selected four drug shops, in four rural market centers and sampled all households in the catchment area (within a 4km radius) of each of these shops."
"The randomization of households was done using a computerized random number assignment algorithm and was stratified by drug shop, by the household's distance to the drug shop (in quartiles) and by the presence of children in the household."
Loss of clusters Low risk Quote: "Only five percent of households surveyed at baseline were not reached at endline, and attrition was balanced across treatment arms."
Incorrect analysis Low risk Comment: Multivariable regression analysis allowing for clustering used (e.g. "If more than one household member got sick simultaneously, we include all concurrent first episodes, and therefore cluster the standard errors in all illness episode regressions at the household level."
Comparability with individually randomised trials Price of AMFm co-paid QAACT < price of AMT (this is useful but not sufficient to determine success)
QAACT price < 150% of the price of the dominant non-QAACT (in most countries this is CQ or SP) and
Price of AMFm co-paid QAACT < price of AMT (this is useful but not sufficient to determine success)
## 34
ACT = artemisinin-based combination therapy AL = artemether-lumefantrine NGO = non-governmental organisation RCT = randomised controlled trial a All studies: no ACT subsidy interventions were implemented in the control sites; malaria diagnosis was predominantly presumptive based on the presence of fever; exemption was granted for ACTs to be provided over the counter in the intervention sites; AL was repackaged in weight specific packs and marked with recommended retail prices to inform consumers the maximum amount they should pay b Specialised drug stores (registered or unregistered pharmacies) and general stores (which sold medicines alongside general household goods) c There was better availability of ACT in the public sector in the control district because of improvements in the procurement and distribution system, and supply by one NGO
## Control
## 37
ACT = artemisinin-based combination therapy CI = confidence interval RCT = randomised controlled trial a Drug shops in population centres were more likely to stock ACTs than those in more remote areas (P < 0.001) . Change in ACT price to patients ACT = artemisinin-based combination therapy AQ = amodiaquine IQR = interquartile range NR = not reported RCT = randomised controlled trial SP = sulphadoxine-pyrimethamine a ACT treatment course: six tablets (for children aged 3-35 months) and 12 tablets (for children aged 36-59 months) -cost estimates based on ACT course for children aged 3-35 months b Di erence between baseline cost per ACT course (control group: US$ 1.08) and follow-up cost per ACT course (intervention group: US$ 0.24) c At follow-up, 95.3% of caregivers in the intervention arm who bought subsidised AL said they purchased it at the recommended retail price of US$ 0.25. Of the eight not paying this price, three paid less than US$ 0.25 and five paid between US$ 0.31 and US$ 1.23 d All age groups: the mean price for ACTs (US$ 0.58) did not di er from the price of SP (US$ 0.67), but was higher than for AQ (US$ 0.48, P < 0.001) e The recommended retail price for an adult course of treatment -US$ 0.47 -was not adhered to (the median price at the endline survey was US$ 1.96) . ACT purchases, sales or market share ACT = artemisinin-based combination therapy AL = artemether-lumefantrine AQ = amodiaquine CI = confidence interval NR = not reported RCT = randomised controlled trial SP = sulphadoxine-pyrimethamine a 95% CI not estimable from the reported data b There was no correlation between the socio-economic status of the consumer and the likelihood of buying ACTs c Purchases of SP and AQ in the intervention districts declined from 68.0% to 51.0% and 26.0% to 11.0% respectively. Purchases of SP in the control district increased from 62.0% to 83.0% while for AQ declined from 33% to 16.0% d Purchases of SP in the intervention districts decreased from 7.0% to 4.0% and remained the same at 9.0% in the control district. Purchases of AQ in the intervention districts declined from 91.0% to 36.0%, and from 91.0% to 36.0% in the control district e The market shares for chloroquine and quinine were 5% and 24% respectively at the end of the pilot; Children less than five years had subsidised ACTs purchased for them more o en than those aged above 5 years Cochrane Database of Systematic Reviews
## Study
## Control
## S o u r c e s o f s u p p o r t
## Internal sources
- KEMRI-Wellcome Trust Research Programme, Kenya.
## External sources
- E ective Health Care Research Consortium, UK.
- Cochrane E ective and Practice and Organisation of Care Group, Norway.
- Evidence-to-Policy initiative (E2Pi), Global Health Group, University of California, USA.
- UKAid (Department for International Development), UK.
## D i f f e r e n c e s b e t w e e n p r o t o c o l a n d r e v i e w
We intended to include only randomised trials, non-randomised trials (with at least two intervention and two control sites), controlled before-a er studies (with at least two intervention and two control sites) and interrupted-time-series studies (Appendix 1). We, however, made a post-hoc decision to include two non-randomised cluster trials (Sabot 2009; Talisuna 2012) that compared intervention sites to only one control site (we downgraded the certainty of evidence in Sabot 2009 and acknowledged possible confounding associated with these designs).
We added adverse e ects (unintended consequences of ACT subsidies) as a secondary outcome.
We did not consider two pre-specified secondary outcomes: availability of alternative antimalarial drugs in all facilities, private and public (including informal outlets); and prices of alternative antimalarial drugs (full adult or child courses; we prioritised direct outcomes presented in the summary-of-findings table).
A number of methods planned in the protocol were not implemented in the review. These methods could be relevant for future updates of this review and are summarised in Appendix 5.
## I n d e x t e r m s
[fig] Figure 1, Figure 2: Illustrative example of the AMFm process Logic framework for evaluating AMFm programme [/fig]
[fig] Figure 3 ;: Moher 2009). Studies initially considered eligible but eventually excluded together with the reasons for exclusions are presented in the Characteristics of excluded studies. [/fig]
[fig] Figure 3: Results of the literature search and studies selected Subsidising artemisinin-based combination therapy in the private retail sector (Review) Copyright © 2016 The Authors. Cochrane Database of Systematic Reviews published by John Wiley & Sons, Ltd. on behalf of The Cochrane Collaboration. [/fig]
[fig] Cohen 2015 ;: Kangwana 2011; Kangwana 2013). The remaining studies were non-randomised cluster trials (Sabot 2009; Talisuna 2012). The studies were conducted in rural districts in Kenya (Cohen 2015; Kangwana 2011; Kangwana 2013), Uganda (Talisuna 2012) and Tanzania (Sabot 2009). One trial enrolled children under five years of age (Kangwana 2011; Kangwana 2013); both adults and children were studied in the other trials. One trial (Kangwana 2011; Kangwana 2013) reported adequate power (80%) for primary outcomes. Study power was not reported in the other three studies. ACT price subsidies were accompanied with supportive interventions (including retail outlet provider training and community awareness campaigns lasting less than a year) in all except one study (Cohen 2015). The subsidy level was 95% in two studies (Kangwana 2011; Kangwana 2013; Talisuna 2012), 80% to 92% in one study (Cohen 2015) and 90% in the remaining study (Sabot 2009). Post-intervention data collection periods were varied: 4 months (Cohen 2015), 12 months (Kangwana 2011; Kangwana 2013; Sabot 2009) and 20 months (Talisuna 2012). Retail outlets comprised specialised and general drug stores in all the studies. Private clinics were included in one study (Talisuna 2012). [/fig]
[table] Table 5: In the first study(Kangwana 2011; Kangwana 2013), ACT subsidy programmes increased the percentage of retail outlets stocking ACTs for children under five years of age by 31.9 percentage points (95% CI 26.3 to 37.5 percentage points; high certainty evidence). In the second study (Sabot 2009), ACT subsidy programmes increased the percentage of retail outlets stocking ACTs for patients of any age by 72.2 percentage points (95% CI 65.0 to 79.3 percentage points; very low quality evidence).Three studies reported data on ACT price outcomes(Kangwana 2011; Kangwana 2013; Sabot 2009; Talisuna 2012; Table 6). In the first study (Kangwana 2011), ACT subsidy programmes decreased the median price for ACT prescribed for children under five years of age by US$ 0.84 (median cost per ACT course without subsidy: US$ 1.08 versus with subsidy: US$ 0.24; high certainty evidence). In the second study (Talisuna 2012), "the maximum recommended retail price was within 10% of the recommended ACT price". In addition, the recommended retail price for an adult course of treatment (US $ 0.47) was not adhered to (the median price at the endline survey was US$ 1.96). In the third study (Sabot 2009), the mean price paid for paediatric ACTs (US$ 0.19) was less than for both SP (US$ 0.51, P = 0.001) and AQ (US$ 0.86, P < 0.001). The mean price for ACTs for any age (US$ 0.58) did not di er from the price of SP (US$ 0.67), but was higher than the price for AQ (US$ 0.48, P < 0.001).Three studies assessed ACT market share outcomes(Kangwana 2011; Kangwana 2013; Sabot 2009; Talisuna 2012; Table 7). In the first study(Kangwana 2011; Kangwana 2013), ACT subsidy programmes increased market share of ACTs among children under five years of age by between 23.6 and 63.0 percentage points (high certainty evidence). In the second study (Talisuna 2012), the market share for ACTs for patients of any age in the intervention group was 43% at baseline and 69% at follow-up (control data not reported). In the third study (Sabot 2009), the market share for ACTs for children under five years of age increased by 8.9% (-0.5% to 18.2%), and 35.3% (29.8% to 40.7%) for patients aged at least 16 years. [/table]
[table] Review: Copyright © 2016 The Authors. Cochrane Database of Systematic Reviews published by John Wiley & Sons, Ltd. on behalf of The Cochrane Collaboration. [/table]
[table] Table 1: Guidelines for success benchmarks at 1 and 2 years a er e ective start date of the AMFm Phase 1 at the country level ACT = artemisinin-based combination therapy; AMT = artemisinin monotherapy AMFm = A ordable Medicines Facility-malaria CQ = chloroquine QAACT = quality-assured artemisinin-based combination therapies SP = sulphadoxine-pyrimethamine a The denominator for ACT use is 'fever episodes in children under age 5' (not 'parasitologically confirmed malaria cases'). The Independent Evaluation relies on national surveys (e.g. demographic and health surveys, multiple indicator cluster surveys, malaria indicator surveys and ACTwatch surveys), which use the denominator 'fever episodes in children under age 5' due to a lack of proper malaria diagnosis in many countries b Price change was the indicator with the weakest empirical basis for setting a 1-year expectation Source: Yamey 2012 [/table]
[table] Table 2: Characteristics of included studies Subsidising artemisinin-based combination therapy in the private retail sector (Review) Copyright © 2016 The Authors. Cochrane Database of Systematic Reviews published by John Wiley & Sons, Ltd. on behalf of The Cochrane Collaboration. [/table]
[table] Table 3: Percentage of children with fever who received ACT on the same day or following day of fever onset ACT = artemisinin-based combination therapy CI = confidence interval NR = not reported OR = odds ratio RCT = randomised controlled trial a There was no correlation between socio-economic status and use of AL (p=0.875) or Tibamal, subsidised AL (p=0.745) b Estimated assuming similar baseline values in control and intervention groups c All age groups: patients in the intervention districts had a six-fold increase in ACT use relative to the control district (95% CI 4.22 to 8.44). Use of ACT was higher in the highest socio-economic status stratum compared to the lowest stratum (OR 2.4, 95% CI 1.72 to 3.35, p<0.001); estimated from available data At 5 years: median price of ACTs in facilities selling ACTs was US$ 4.04 (ACTs free in public facilities); ACTs 11.3 times more expensive than the most common antimalarial in outlets selling ACTs [/table]
[table] Table 8: Use of older antimalarials CI = confidence interval NR = not reported OR = odds ratio RCT = randomised controlled trial a amodiaquine, suphadoxine-pyrimethamine and quinine [/table]
|
A web-based interactive framework to assist in the prioritization of disease candidate genes in whole-exome sequencing studies
Whole-exome sequencing has become a fundamental tool for the discovery of disease-related genes of familial diseases and the identification of somatic driver variants in cancer. However, finding the causal mutation among the enormous background of individual variability in a small number of samples is still a big challenge. Here we describe a web-based tool, BiERapp, which efficiently helps in the identification of causative variants in family and sporadic genetic diseases. The program reads lists of predicted variants (nucleotide substitutions and indels) in affected individuals or tumor samples and controls. In family studies, different modes of inheritance can easily be defined to filter out variants that do not segregate with the disease along the family. Moreover, BiERapp integrates additional information such as allelic frequencies in the general population and the most popular damaging scores to further narrow down the number of putative variants in successive filtering steps. BiERapp provides an interactive and user-friendly interface that implements the filtering strategy used in the context of a large-scale genomic project carried out by the Spanish Network for Research in Rare Diseases (CIBERER) in which more than 800 exomes have been analyzed. BiERapp is freely available at: http://bierapp.babelomics.org/
# Introduction
Recent advances in high-throughput sequencing technologies have made it possible to sequence whole genomes or exomes at unprecedented speeds and low costs. In particular, targeted sequencing of exomes has been extensively and successfully used to discover disease genes in Mendelian disorders [bib_ref] Exome sequencing identifies the cause of a mendelian disorder, Ng [/bib_ref] or in cancer [bib_ref] Cancer genome landscapes, Vogelstein [/bib_ref]. However, with more than 30 000 variants found per exome (1), finding disease-causing genes is a cumbersome, time-consuming task that often requires intensive human intervention [bib_ref] Needles in stacks of needles: finding disease-causal variants in a wealth of..., Cooper [/bib_ref].
In spite of the obvious need for tools that facilitate the gene prioritization process, there are no many open solutions currently available [bib_ref] Sequencing studies in human genetics: design and interpretation, Goldstein [/bib_ref]. Most of the available tools cover the primary analysis (QC, alignment and variant calling) (7-9) that ends up in a list of variants found in sequencing experiments (VCF file) that can be annotated with different programs, such as VARIANT [bib_ref] VARIANT: Command Line, Web service and Web interface for fast and accurate..., Medina [/bib_ref] , ANNOVAR [bib_ref] ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data, Wang [/bib_ref] , etc.
In the case of inherited diseases or de novo syndromes, the availability of sequencing data of parents, siblings or close relatives can significantly help in the process of finding candidate disease genes. Actually, more sophisticated tools can use such information to help in the detection of diseasecausing variants segregating along family pedigrees [bib_ref] Rare variant detection using family-based sequencing analysis, Peng [/bib_ref] [bib_ref] A likelihood-based framework for variant calling and de novo mutation detection in..., Li [/bib_ref] or somatic mutations in cancer [bib_ref] VarScan 2: somatic mutation and copy number alteration discovery in cancer by..., Koboldt [/bib_ref]. These tools increase the precision of the calling process but lack, in some cases, userfriendliness and fail to provide other filtering steps. The only tool that enables more filtering steps, KGGSeq [bib_ref] A comprehensive framework for prioritizing variants in exome sequencing studies of Mendelian..., Li [/bib_ref] , is a command line application.
BiERapp fills the gap that leads from the list of predicted variants (VCF file) to the final candidate disease-gene list by providing an interactive, web-based, easy-to-use framework. The tool allows for the consecutive application of filters that include segregation in familial cases (with different inheritance modes that can be easily defined), allelic frequencies in the general population, mutational consequences and other that narrow down the number of putative variants to a small number of promising candidate variants.
# Materials and methods
## Input
BiERapp accepts standard VCF formats, typically a multi-sample VCF corresponding to several samples of one or several families. Sporadic patients or case-control studies can also be analyzed within the same framework.
The 'Data Upload' entry of the left menu of the main screen (see [fig_ref] Figure 1: The main screen of BiERapp and the filtering panels [/fig_ref] allows uploading the data from the user's local disk. Using the 'Upload local file' button, the user can browse their local disk and load the VCF file. In the dialog window that pops up, the user has some options to identify the data and to associate some information with them. The selected local file is uploaded to the server, where several VCF files can be stored. Pressing the run button the data set goes to the 'jobs' panel, on the right, where it is indexed. This process may take several minutes, depending on the data size. For example, a VCF file containing some 40 000 variants may take about 10 min. Supplementary [fig_ref] Figure 1: The main screen of BiERapp and the filtering panels [/fig_ref] shows a linear relationship between indexing runtimes and the number of variants, with different slopes for different number of exomes. The figure shows a representation of indexing runtimes for different numbers of exomes (from 1 to 20) and different numbers of variants (up to 100 000). Once the job is finished, the user can click on it and the prioritization by applying successive filtering steps can start. Several jobs can independently be run and invoked for analysis.
## Data sources
The information used for the annotation of the position is stored in CellBase [bib_ref] CellBase, a comprehensive collection of RESTful web services for retrieving relevant biological..., Bleda [/bib_ref] , which collects it from different sources. The identifiers of single nucleotide polymorphism (SNP) are extracted from dbSNP [bib_ref] ) dbSNP: the NCBI database of genetic variation, Sherry [/bib_ref]. The consequence types of the variants and their predicted pathologic effect, according Polyphen [bib_ref] Human non-synonymous SNPs: server and survey, Ramensky [/bib_ref] or SIFT [bib_ref] Predicting the effects of coding non-synonymous variants on protein function using the..., Kumar [/bib_ref] indexes, are extracted from Ensembl. The Minor Allele Frequencies (MAFs) of the variants are calculated for the populations derived from the 1000 genomes [bib_ref] A map of human genome variation from population-scale sequencing, Durbin [/bib_ref] and Exome Sequencing Project (ESP) (23) studies. In the case of 1000 genomes, the VCF files are downloaded from the server (http://www. 1000genomes.org/ftpsearch) and the proportions of the different genotypes are calculated and the allelic frequencies derived from them. In the ESP, the genotype counts were directly available in the server (http://evs.gs.washington.edu/ EVS/) and were used to derive MAFs. The disease phenotype from HGMD (24), ClinVar [bib_ref] ClinVar: public archive of relationships among sequence variation and human phenotype, Landrum [/bib_ref] and UNIPROTdatabases is taken from Ensembl as well.
## Prioritization of variants by successive filtering
Each prioritization ('job') has three associated screens that facilitate the filtering steps. The first one, the 'Summary' tab, displays a statistic of the data set analyzed, containing the samples analyzed, the number and types of variants found and its distribution according to consequence types. The second screen, in the 'Variants and effect' tab, is the actual filtering tool, and the third one, the 'Genome view' tab, offers a representation of the selected variants within the genomic context provided by an embedded version of the Genome Maps tool [bib_ref] Genome Maps, a new generation genome browser, Medina [/bib_ref].
The prioritization of variants is conducted by means of a consecutive filtering strategy in which different filters are applied to reduce the number of potential disease variants. The filters that can be applied are the following.
r Segregation filter: This filter allows specifying the distribution of alleles that are compatible with the pedigree analyzed, given the inheritance model of the disease. [fig_ref] Figure 1: The main screen of BiERapp and the filtering panels [/fig_ref] shows an example with a very simple family 'pedigree'. If the disease is autosomal recessive then: both parents must be carriers and have the alleles in a configuration 0/1 (meaning that one of the alleles is the reference allele and the other one is the alternative allele), the affected individual must have an allelic configuration of 1/1 (that is, homozygote for the alternative allele, which is the potential causative agent) and the unaffected individual could either be 1/0 or 0/0 (that is, heterozygote or homozygote for the reference allele, respectively). Any pedigree and inheritance model, including incomplete penetrance, can be specified in the interface [fig_ref] Figure 1: The main screen of BiERapp and the filtering panels [/fig_ref] using very simple rules in an intuitive manner. Obviously, the application of the filter discards Mendelian inconsistencies.
r The segregation filter can also be used to analyze casecontrol experimental designs by setting the configuration of alleles in a way that differentiates among them (e.g. 1/1 for cases versus 1/0 or 0/0 for controls in the case of a mutation that causes loss of function).
r Since sequencing errors can happen, or some of the cases could have a different mutation, the filter can accept a certain degree of uncertainty in the application of the filters. Thus, the data available for some samples could be missing for the filtered position. This maximum number of missing values accepted can be indicated in the corresponding box [fig_ref] Figure 1: The main screen of BiERapp and the filtering panels [/fig_ref]. When N missing values are accepted, the positions that either fulfill the zygosity criteria or are missing are first collected. Then only those positions having N or fewer samples with missing values are displayed. r Consequence type filter: As a first option, variants with a predicted severe effect are the best candidates. Therefore, 'non-synonymous' and 'stop lost' are the initial preferred consequence types.
r Allelic frequency filter: This is another quite useful filter that can be used to discard variants with a relatively high MAF in the population. These variants are unlikely to be causative of many hereditary disorders. MAFs are obtained from the 1000 genomes [bib_ref] A map of human genome variation from population-scale sequencing, Durbin [/bib_ref] and ESP (23) projects.
r Regions filter: Should previous information on regions associated with the disease be available from linkage disequilibrium studies, BiERapp allows focusing on these particular regions, which can be specified in the corresponding box.
r Genes filter: If some genes are of special interest, the analysis can be focused on them by specifying their names in the corresponding box.
When the filter is applied, a selection of filtered variants appears in the 'Variant info' panel. Each line corre- sponds to a variant for which the following information is displayed: (i) genomic position (chromosome:position); (ii) allelic change (reference allele > alternative allele); (iii) the name of the gene affected by the variant; (iv) the allelic composition of each sample analyzed (0/0, 0/1 and 1/1 for the homozygote reference allele, the heterozygote and the homozygote alternative allele, respectively; ./. accounts for low quality or low coverage positions); (v) the SNP identifier in dbSNP [bib_ref] ) dbSNP: the NCBI database of genetic variation, Sherry [/bib_ref] if the variant is an already known SNP; (vi) the MAF in the population derived from the 1000 genomes [bib_ref] A map of human genome variation from population-scale sequencing, Durbin [/bib_ref] and ESP (23) studies; (vii) the consequence type of the variant; (viii) the predicted pathologic effect of the variants according to Polyphen [bib_ref] Human non-synonymous SNPs: server and survey, Ramensky [/bib_ref] or (ix) SIFT [bib_ref] Predicting the effects of coding non-synonymous variants on protein function using the..., Kumar [/bib_ref] indexes (if the variant has more than one pathologic consequence, the most deleterious value of the indexes is displayed here); and the phenotype extracted from CellBase (17) as annotated in HGMD [bib_ref] The Human Gene Mutation Database (HGMD) and its exploitation in the fields..., Stenson [/bib_ref] , ClinVar [bib_ref] ClinVar: public archive of relationships among sequence variation and human phenotype, Landrum [/bib_ref] and UNIPROT. Columns can be customized by hiding or rearranging them by directly clicking on them or using the button 'columns' in the lower right corner of the panel.
A second panel displays the effect that the variant selected in the first panel has over the different genomic fea-tures in which it is located (that can be more than one). It often occurs that the same variant affects different transcripts, with different effects, and also to other elements such as regulatory motifs, splicing motifs, etc. This panel provides exhaustive information on all the aspects of the possible effect of the variant and includes the following data: chromosomal position, SNP ID (if any), consequence type, amino acid change (if the feature is an exon and applies), the Ensembl gene name, the transcript ID (including a link to the Ensembl), the feature ID, the feature name, type and biotype.
If no variants compatible with the disease are found or further validations demonstrate that none of the selected variants were associated with the disease then the filters can be relaxed to increase the number of possible candidates. New prioritizations can be interactively generated by changing the filters and pressing the 'search' button in the 'Variants and effect' tab.
Finally, a third tab displays the selected variant within the genomic context using the Genome Maps [bib_ref] Genome Maps, a new generation genome browser, Medina [/bib_ref] genome viewer, which is embedded in the application. Genome Maps provides a contextual view of the variant position within the genome, highlighting all the relevant features around (transcripts, genes, SNPs, etc.).
Filtering sessions are deleted once the web page is closed. Alternatively, there is a possibility of registering and logging in on a private session. In this case, all the analyses done are kept in a user's account in the server.
## Output
Data can be exported in comma delimited CSV format, which can be imported by any spreadsheet. In the 'Variants and effect' tab, the 'variant info' panel has in its lower right corner a button to export data. When clicking on it, a window pops up where the user can select the columns to save and the CSV file can be downloaded to the local disk.
## Technical details
BiERapp is an open source tool based on HTML5 and JavaScript. The application user interface has been developed in javascript with the Ext JS and the Bootstrap framework. BiERapp uses a fast and optimized indexing and annotating system based in SQlite for queries. All the filtering operations are carried out locally, in the user's browser. The relevant information on genes, variants, features, etc. used for the prioritization is remotely stored (and kept updated) in CellBase [bib_ref] CellBase, a comprehensive collection of RESTful web services for retrieving relevant biological..., Bleda [/bib_ref] and is provided through highly efficient web services.
In order to scale up and improve the database performance, a second implementation has been developed using MongoDB database. MongoDB is a distributed and scalable high-performance database. This implementation shows a much higher runtime performance and can scale up to Terabytes of data. BiERapp fetches data through Java RESTful web services which can query these two possible database implementations, making transparent for the user how the application is storing the data.
BiERapp makes an intensive use of new web technologies and standards, therefore only new browsers are fully supported. These include: Chrome 14+, Firefox 7+, Safari 5+, Internet Explorer 11+ and Opera 11+. Older browser like Chrome13−, Firefox 5− or Internet Explorer 9 may yield errors. Internet Explorer 6 and 7 are no supported.
# Discussion
BiERapp is an extensively tested tool, which has been used during the last year by the BiER team (Bioinformatics for Rare Diseases Team; http://www.ciberer.es/bier). More than 800 exomes of patients of more than 70 different inherited pathologies, produced by the Spanish Network for Research in Rare Diseases (CIBERER, http://www. ciberer.es) and the Medical Genome Project (MGP; http: //www.medicalgenomeproject.com), have been analyzed using BiERapp. Recent publications include the discovery of two new mutations in the BCKDK gene, responsible of a neurobehavioral deficit in pediatric patients [bib_ref] Two Novel Mutations in the BCKDK Gene (Branched-Chain Keto-Acid Dehydrogenase Kinase) are..., Garcia-Cazorla [/bib_ref] , new mutations in different genes causing inherited retinal dystrophies [bib_ref] Web Server issue W93 genetic and high-throughput strategies for molecular diagnosis of..., De Castro-Miro [/bib_ref] [bib_ref] Whole-exome sequencing identifies novel compound heterozygous mutations in USH2A in Spanish patients..., Mendez-Vidal [/bib_ref] [bib_ref] Mutation screening of multiple genes in Spanish patients with autosomal recessive retinitis..., Gonzalez-Del Pozo [/bib_ref] [bib_ref] Mutation spectrum of EYS in Spanish patients with autosomal recessive retinitis pigmentosa, Barragan [/bib_ref] and metabolic diseases [bib_ref] Exome sequencing identifies a new mutation in SERAC1 in a patient with..., Tort [/bib_ref].
There are several programs available in which family information is used to improve the variant calling process like FamSeq [bib_ref] Rare variant detection using family-based sequencing analysis, Peng [/bib_ref] , PolyMutt [bib_ref] A likelihood-based framework for variant calling and de novo mutation detection in..., Li [/bib_ref] or the recently published Vari-antMaster (34) which claim to be quite sensitive and especially devised to detect de novo variants. Other tools are also sophisticated variant callers specific for finding somatic variation in cancer [bib_ref] VarScan 2: somatic mutation and copy number alteration discovery in cancer by..., Koboldt [/bib_ref]. KGGSeq (15) is a java application with a filtering philosophy similar to those in BiERapp. However, the inheritance pattern filter seems to be too rigid and offers only a limited number of scenarios. None of the mentioned applications is a web server.
Apart from free applications there are several commercial solutions available. Ingenuity's Variant Analysis (http://www.ingenuity.com/products/variant-analysis) and Golden Helix's SNP & Variation Suite (http://www.goldenhelix.com/SNP Variation/DNA-Seq Analysis Package/index.html) offer a sequential filtering strategy similar to BiERapp. The second one is a stand-alone application that offers a filtering step similar to what we have implemented in BiERApp.
The proposed web-based interactive framework has great potential to detect disease-related variants in familial diseases as demonstrated by its successful use in the CIBERER and MGP initiatives above mentioned. To our knowledge, BiERapp is the first free web tool that provides the possibility of applying a consecutive filtering approach to variants coming from a whole-exome sequencing study with this level of interactivity. The program manages Mendelian inheritance modes by providing an intuitive filter that allows reproducing any familial pedigree with any inheritance model and allows selecting variants (and genes with deleterious variants) that segregate with the disease in the family. The use of the filters is interactive and the results are almost instantaneously displayed in a panel that includes the genes affected, the variants, the consequence types, allelic frequencies in 1000 genomes (as a whole or in four groups with ancestors of different geographical origin: European, Asian, African and American) and ESP, as well as other parameters of interest. A known cause of generation of false positives and false negatives is the existence of regions with poor coverage or low quality in which variants are not reported in some of the individuals analyzed. Unlike any other tool, BiERapp manages efficiently these missing regions (if annotated in the VCF files) and consequently allows for variant filtering through pedigrees containing noisy or incomplete data.
When enough samples are available, complete family 'pedigrees' can be used and the experimental design is reasonable, the final number of candidates is usually a small figure and no more prioritization steps are necessary. In this case, the necessary subsequent experimental validation of only a small number of candidates contributes to the optimization of resources (time and budget) in the disease gene discovery process. However, this is not always the case. Often, the availability of other family members is not guaranteed or the own nature of the disease (rare diseases) precludes obtaining the necessary number of samples. Nevertheless, even in this case, the application of all the possible filters reduces in orders of magnitude the number of possible candidates. For example, most Mendelian diseases are caused by rare variants thus filtering out candidate variants present in control populations is of crucial importance [bib_ref] Evolution and functional impact of rare coding variation from deep sequencing of..., Tennessen [/bib_ref].
If still a large number of candidate genes are obtained, other prioritization methods that make use of other gene properties (e.g. physical, genetic or functional relationships among them or to other known disease genes) can be used [bib_ref] SNOW, a web-based tool for the statistical analysis of protein-protein interaction networks, Minguez [/bib_ref] [bib_ref] Genome-wide prioritization of disease genes and identification of disease-disease associations from an..., Linghu [/bib_ref] [bib_ref] Computational tools for prioritizing candidate genes: boosting disease gene discovery, Moreau [/bib_ref]. As future improvement, we plan to provide a functional layer for further knowledge-based prioritization by connecting BiERapp to several methods implemented in our functional profiling tool Babelomics [bib_ref] Babelomics: an integrative platform for the analysis of transcriptomics, proteomics and genomic..., Medina [/bib_ref].
## Supplementary data
Supplementary Data are available at NAR Online.
[fig] Figure 1: The main screen of BiERapp and the filtering panels. (A) Example of family pedigree with an affected descendant (NA19600), two carrier parents and an unaffected descendant. (B) The segregation filter that allows easily defining any inheritance mode (see the text for an explanation). [/fig]
|
Chemical conversion of cisplatin and carboplatin with histidine in a model protein crystallized under sodium iodide conditions
Cisplatin and carboplatin are platinum anticancer agents that are used to treat a variety of cancers. Previous X-ray crystallographic studies of carboplatin binding to histidine in hen egg-white lysozyme (HEWL) showed a partial chemical conversion of carboplatin to cisplatin owing to the high sodium chloride concentration used in the crystallization conditions. Also, the cocrystallization of HEWL with carboplatin in sodium bromide conditions resulted in the partial conversion of carboplatin to the transbromoplatin form, with a portion of the cyclobutanedicarboxylate (CBDC) moiety still present. The results of the co-crystallization of HEWL with cisplatin or carboplatin in sodium iodide conditions are now reported in order to determine whether the cisplatin and carboplatin converted to the iodo form, and whether this took place in a similar way to the partial conversion of carboplatin to cisplatin in NaCl conditions or to transbromoplatin in NaBr conditions as seen previously. It is reported here that a partial chemical transformation has taken place to a transplatin form for both ligands. The NaI-grown crystals belonged to the monoclinic space group P2 1 with two molecules in the asymmetric unit. The chemically transformed cisplatin and carboplatin bind to both His15 residues, i.e. in each asymmetric unit. The binding is only at the N atom of His15. A third platinum species is also seen in both conditions bound in a crevice between symmetry-related molecules. Here, the platinum is bound to three I atoms identified based on their anomalous difference electron densities and their refined occupancies, with the fourth bound atom being a Cl atom (in the cisplatin case) or a portion of the CBDC moiety (in the carboplatin case).
Cisplatin and carboplatin are platinum anticancer agents that are used to treat a variety of cancers. Previous X-ray crystallographic studies of carboplatin binding to histidine in hen egg-white lysozyme (HEWL) showed a partial chemical conversion of carboplatin to cisplatin owing to the high sodium chloride concentration used in the crystallization conditions. Also, the cocrystallization of HEWL with carboplatin in sodium bromide conditions resulted in the partial conversion of carboplatin to the transbromoplatin form, with a portion of the cyclobutanedicarboxylate (CBDC) moiety still present. The results of the co-crystallization of HEWL with cisplatin or carboplatin in sodium iodide conditions are now reported in order to determine whether the cisplatin and carboplatin converted to the iodo form, and whether this took place in a similar way to the partial conversion of carboplatin to cisplatin in NaCl conditions or to transbromoplatin in NaBr conditions as seen previously. It is reported here that a partial chemical transformation has taken place to a transplatin form for both ligands. The NaI-grown crystals belonged to the monoclinic space group P2 1 with two molecules in the asymmetric unit. The chemically transformed cisplatin and carboplatin bind to both His15 residues, i.e. in each asymmetric unit. The binding is only at the N atom of His15. A third platinum species is also seen in both conditions bound in a crevice between symmetry-related molecules. Here, the platinum is bound to three I atoms identified based on their anomalous difference electron densities and their refined occupancies, with the fourth bound atom being a Cl atom (in the cisplatin case) or a portion of the CBDC moiety (in the carboplatin case).
# Introduction
Cisplatin and carboplatin are platinum anticancer drugs that have long been used as cytostatics (http://www.cancerresearchuk.org/ cancer-help/about-cancer/cancer-questions/what-are-cytostatic-drugs) in the fight against cancer by targeting DNA. Our previous X-ray crystallographic studies of carboplatin with HEWL, a model protein, have shown partial chemical conversion to cisplatin owing to the high sodium chloride conditions used in the crystallization mixture, as confirmed by the presence of extra anomalous difference density in the binding site that is not attributable to the Pt atom. Subsequently, we replaced sodium chloride with sodium bromide in the crystallization mixture to more easily observe the conversion of carboplatin to the bromo form. In this condition, the observed chemical transformation which took place was to the transbromoplatin form, and contrasted with the sodium chloride case where cisplatin was seen. Since a portion of the cyclobutanedicarboxylate (CBDC) moiety was still visible in the His15 ligand-binding site under the sodium bromide conditions, this confirmed that the chemical conversion of carboplatin to transbromoplatin was partial.
In the study reported here, cisplatin or carboplatin was cocrystallized with HEWL under sodium iodide conditions. The motivation for this new study is twofold. Firstly, to again confirm whether a chemical conversion occurs, and specifically whether it is to the trans-platinated or cis-platinated form and whether it is a partial or a full chemical conversion. Secondly, if an iodinated platinated species is observed then it would suggest the prospect of undertaking a dual K-edge (platinum and iodine) targeted radiation therapy experiment against a tumour. Cisplatin administration has previously been studied in the case of radiation therapy using a synchrotron source and the Pt K edge alone; see also http:// www.esrf.eu/UsersAndScience/Publications/Highlights/2004/Imaging/ Ima10/). Sodium iodide has been used previously to crystallize native HEWL and in these conditions HEWL crystallizes in space group P2 1 with two protein molecules in the asymmetric unit.
# Methods
## Crystallization conditions
Co-crystallization of HEWL with cisplatin in NaI solution was carried out under similar conditions as published inandbut with 20 mg HEWL co-crystallized with 1.2 mg cisplatin in 75 ml DMSO, 462.5 ml 0.1 M sodium acetate, 462.5 ml 1 M NaI solution. The conditions were identical for carboplatin but used 1.4 mg carboplatin.
## X-ray diffraction data collection, structure solution and model refinement
A crystal of HEWL with cisplatin or carboplatin was scooped into a loop with silicone oil used as the cryoprotectant. All X-ray diffraction (XRD) data were measured on a Bruker APEX II homesource diffractometer at an X-ray wavelength of 1.5418 Å at a fixed temperature of 100 K [fig_ref] Table 1: X-ray crystallographic data and final protein plus ligand model-refinement statistics for both... [/fig_ref] , using an XRD data-collection strategy to obtain generally good data sets, i.e. high completeness of unique data, high anomalous difference completeness and good data redundancy. XRD data from both crystals were processed using the SAINT software package (v.2; Bruker AXS Inc., Madison, WI, USA).
The crystal structures were solved using molecular replacement with followed by rigid-body refinement with REFMAC5 from CCP4using the reported monoclinic lysozyme structure with PDB code 1lkr as the molecular search model. Amplitude twin refinement was carried out for the cisplatin data set. Anomalous difference maps were generated using calculated phases with the ligands omitted from the model. These maps allowed a check for and identification of the I-atom positions. Model building, adjustment and restrained refinement were carried out using Coot and REFMAC5in CCP4. Ligand-binding occupancies were calculated using SHELXTL. The platinum-to-ligand distances were not restrained during refinement. The crystallographic and molecular model-refinement parameters are summarized in [fig_ref] Table 1: X-ray crystallographic data and final protein plus ligand model-refinement statistics for both... [/fig_ref]
# Results
## Cisplatin study
Ligand binding is seen at His15 residue of both molecules A and B of HEWL in this monoclinic crystal form, i.e. with two molecules per asymmetric unit. In molecule A platinum binding is only seen at the N binding site , with an anomalous difference electrondensity peak height of 9.8 and an occupancy value of 80% [fig_ref] Table 2: Occupancy values of the Pt and I atoms for both the cisplatin... [/fig_ref]. Besides the platinum peak, there are two large anomalous difference electron-density peaks of 11.1 and 8.8 in the trans positions which are readily assignable as I atoms at distances of 2.6 Å (AE0.1 Å ) from .4 ‡ † The PDB validation report includes phenix.xtriage statistics; subsequent use of DETWIN in CCP4 suggested that the data were 6% twinned. To our knowledge, no twinning has been reported for this form of HEWL in NaI conditions before. Turning on amplitude twin refinement in REFMAC improved the R/R free from 19.7/27.5% to 19.7/ 27.1%. The electron-density maps at the His binding sites as well as in general were the same with and without using twin refinement, presumably owing to the rather small twin fraction. ‡ The disallowed residues are Asp18, Ser72 and Arg73 and are part of a loop/ turn region.
## Figure 1
Binding to His15 shows a chemically transformed cisplatin, namely transiodoplatin. (a, b) The molecule A binding site shown in two different views. The 2F o À F c electron-density map (blue) is contoured at 1.5 r.m.s. and the anomalous difference electron-density map (orange) is contoured at 3. The Pt atom is shown in purple, the iodines are in yellow, the chlorine is in grey, C atoms are in green and N atoms are in blue.
the Pt atom, confirming that cisplatin has converted to the trans iodoplatinated form (transiodoplatin). The angle between these three atoms is 176 , which is close to linearity. A third 2F o À F c electrondensity peak is seen bound to the Pt atom. However, no anomalous difference electron density is observed here and so a third I atom is ruled out. Modelling in an N atom from cisplatin gave a B factor of 2.0 Å 2 , which is physically unrealistic, and therefore a Cl atom was modelled in. This Cl atom has an occupancy of 100% as calculated by SHELX and a quite reasonable B factor of 21.3 Å 2 . The distance between the platinum and this chlorine is 2.4 Å (AE0.1 Å ) and can be compared with the usual platinum-chlorine distance of 2.35 Å ; as this distance is well within the error of the bond-distance estimate (0.1 Å ), this atom assignment seems reasonable. In molecule B, the identification of the compound bound to the N atom of His15 (Supplementary 1 ) is more difficult than in molecule A described above and the figure is given in the Supporting Information for completeness. Platinum binding is again only seen at the N binding site , with an anomalous difference electron-density peak height of 6.7 and an occupancy of 67% [fig_ref] Table 2: Occupancy values of the Pt and I atoms for both the cisplatin... [/fig_ref]. There are also two other anomalous difference electron-density peaks in the binding site (7.2 and 4) in the trans position to the Pt atom and these are therefore assigned as I atoms bound to the platinum, each at a distance of 2.6 Å (AE0.2 Å ). The angle between these three atoms is 169 , showing signs of some distortion from linearity. In molecule B there are also two extra 2F o À F c density peaks close to the Pt and I atoms. Modelling in an N atom from cisplatin again gave a B factor of 2.0 Å 2 . Owing to this, a mixture of cisplatin and transiodoplatin appears to be bound to this N atom. The distance between platinum and chlorine 1 is 2.4 Å (AE0.3 Å ) and the platinum-chlorine 2 distance is 2.8 Å (AE0.1 Å ). The occupancies of each molecule as a whole, estimated using SHELX, gave values of 51% for the transiodoplatin molecule and 49% for the cisplatin molecule. A third 2F o À F c electron-density peak is also seen, with no anomalous difference electron density observed at this position. An Na ion with 100% occupancy has been modelled in with a B factor of 33 Å 2 .
At the His15 N " binding sites in molecules A and B, an anomalous difference electron-density peak of 4.1 and 4.0 is seen, respectively, at 3.9 (AE0.2) and 5.0 Å (AE0.2 Å ) from the N " atom, too distant to be a Pt atom. This peak is instead assigned as an I atom in both cases, which is 3.9 Å (AE0.2 Å ) from the NH backbone group of Ile88, which is in the correct distance range for a halogen hydrogen bond. This is the same situation as previously seen in the NaBr crystallization conditions case, where a Br atom was assigned at this position.
## Carboplatin study
Ligand binding is seen at the His15 N atoms in both molecule A and molecule B. In molecule A, a platinum peak with an anomalous difference electron density of 9.8 is seen which refines to an occupancy of 77% and [fig_ref] Table 2: Occupancy values of the Pt and I atoms for both the cisplatin... [/fig_ref]. Two further anomalous difference electron-density peaks are seen (10.8 and 7.8) in the trans position to the Pt atom and these are assigned as I atoms bound to the Pt atom with a distance of 2.6 Å (AE0.3 Å ) . A third 2F o À F c electron-density peak is seen at a binding distance from the Pt atom, but no anomalous difference electron density is observed here. Owing to the lack of anomalous difference density, a portion of the CBDC moiety was modelled in. Thus, similar to the carboplatin in NaBr conditions, a portion of the CBDC moiety must still be present at this binding position.
In molecule B, a platinum anomalous difference electron-density peak of 12.2 is observed which refines to an occupancy of 68% and [fig_ref] Table 2: Occupancy values of the Pt and I atoms for both the cisplatin... [/fig_ref]. Two anomalous difference density peaks are seen (12.1 and 11.3) in trans positions to the Pt atom and these are assigned as I atoms bound to the Pt atom at a distance of 2.6 Å (AE0.2 Å ). A third 2F o À F c electron-density peak is seen at a binding distance to the Pt atom, with a weak anomalous difference electrondensity peak of 2.6. Owing to the weakness of the peak it was not assigned to an atom type, but the 2F o À F c density is more elongated than in the trans iodo positions at this site and could correspond to a mixture of an I atom and a portion of the CBDC moiety . Modelling in a mixture of an I atom and a portion of the CBDC moiety gives occupancies of 52% for the I atom and 48% for the CBDC moiety. Thus, the carboplatin has partially converted to the trans iodo-platinated form, with some evidence that a portion of the carboplatin molecule is still present. Molecule B is shown in and molecule A in as they are very similar and the electron density around the carboplatin is slightly better for molecule B.
At the N " binding site in molecules A and B anomalous difference electron-density peaks of 3.4 and 5.5 are seen 5.1 (AE0.4) and 5.5 Å (AE0.2 Å ) from the N " atom, too distant and too weak to be assignable as a Pt atom. This peak is assigned as an I atom in both cases which is 3.5 Å (AE0.4 Å ) and 3.9 Å (AE0.2 Å ) from the NH backbone group of Ile88, which is in the correct distance range for a halogen hydrogen bond.
## A third pt ligand-binding site in a crevice between two symmetry-related molecules
For both the cisplatin and carboplatin HEWL NaI crystal studies, a third platinum binding site is seen where a PtI 3 moiety is identified bound in a crevice between two symmetry-related molecules near Asn106, Arg112, Pro70, Asn65 and Asn103 of molecule A, and Arg128, Cys6, Glu7 and Ala10 of a symmetry-related molecule. However, this binding is not seen at the same residues in molecule B.
In the cisplatin case, three anomalous difference electron-density peaks of 9.4, 8.2 and 7.8 are seen [fig_ref] Table 2: Occupancy values of the Pt and I atoms for both the cisplatin... [/fig_ref] at distances of 2.7 (AE0.1), 2.5 (AE0.1) and 2.5 Å (AE0.1 Å ) from the Pt atom, respectively. A fourth 2F o À F c electron-density peak is found at a distance of 2.4 Å (AE 0.2 Å ) from the Pt atom. A Cl atom is assigned at this fourth position. In the carboplatin study, three anomalous difference electron-density peaks of 5.4, 5.9 and 9.2 are seen [fig_ref] Table 2: Occupancy values of the Pt and I atoms for both the cisplatin... [/fig_ref] at distances of 2.7 (AE0.3), 2.5 (AE0.3) and 2.7 Å (AE0.3 Å ) from the Pt atom, respectively. A fourth 2F o À F c electron-density peak within binding distance of the Pt atom has elongated density. We interpret this as a portion of the CBDC moiety of carboplatin bound at this fourth position.
# Discussion
Our studies have investigated whether co-crystallizing cisplatin or carboplatin with HEWL in NaI conditions, rather than using NaCl, would result in partial or full chemical conversion of these compounds to the iodo form. Previous results with carboplatin in NaCl crystallization conditions showed partial conversion to cisplatin, whereas in sodium bromide conditions the carboplatin was partially converted to the trans bromo form with a portion of the CBDC moiety still being present. Obviously, cisplatin with HEWL under sodium chloride conditions would simply preserve chlorinated platinum in the cisplatin. The results presented here for both cisplatin and carboplatin indeed showed partial conversion to the trans iodo form with a portion of a Cl or CBDC moiety still seen in the binding site , which is similar to the results using NaBr as the crystallization high-salt component.describe carboplatin binding to HEWL in NaBr crystallization conditions as well as in non-NaCl and non-NaBr conditions, the latter being in order to see a chemically unmodified carboplatin. A study of cisplatin binding to HEWL in NaBr conditions has also been carried out, and for completeness those results are given in the Supporting Information to this paper. The results are very similar to the carboplatin study in NaBr conditions, in which the cisplatin is also seen to be converted to a transbromoplatin form with a Cl atom also bound to the platinum centre at the N binding site. Similar to the carboplatin study, the N " binding site shows less detail, with a weakly occupied Pt atom (20%) bound to a Br atom.
Previously, under NaCl conditions at pH 4.7, we have seen clear binding at both the N and N " sites of His15. This was attributed to the chlorine extracting a His15 H atom, creating an imidazolyl ion. In the bromo casebinding at both sides was still seen, but with a much weaker occupancy (up to 20%) for the N " binding site. We rationalize this as being owing to bromine ions being weaker than chloride ions at extracting A PtI 3 X complex bound in a crevice between two symmetry-related molecules. (a, b) Complex derived from cisplatin with a Pt atom bound to three I atoms and a Cl atom bound near Arg128, Cys6, Glu7 and Ala10 of molecule A and Asn106, Arg112, Pro70 and Asn65 of a symmetry-related molecule. (c, d) Complex derived from carboplatin with platinum bound to three I atoms and a portion of the CBDC moiety bound near Asn106, Arg112, Pro70, Asn65 of molecule A and Arg128, Cys6, Glu7 and Ala10 of a symmetry-related molecule (a). The 2F o À F c electron-density map (blue) is contoured at 1.5 and the anomalous difference electron-density map (orange) is contoured at 3. The Pt atom is shown in purple, the iodines are in yellow, the chlorine is grey, C atoms are green and O atoms are red. the last His H atom to generate the imidazolyl ion. Another possibility is tautomeric forms; seefor a full discussion. Where binding is seen at only one histidine N atom we are obviously not seeing a tautomeric pair or an imidazolyl anion. In the iodo case, the hydrogen-abstracting effectiveness is worse than for bromide ions and this is a likely reason why the binding of a platinum ligand is seen only at one His15 N atom, at least at pH 4.7, in the NaI case. A further difference between the NaI conditions and the NaCl or NaBr conditions was the fact that the crystals are monoclinic, rather than tetragonal, and thereby had two protein molecules in the asymmetric unit. In both monoclinic asymmetric units we note that binding only occurs to the N atom of His15.
Recently,soaked PtI 2 (NH 3 ) 2 into pre-grown HEWL crystals and the Pt atom was seen to be bound to the N atom of His15 but, as they described it, showed 'peculiar features' involving the presence of three peaks of anomalous electron density close to a Pt atom suggesting the presence of two alternative modes of binding of the [PtI 2 NH 3 ] moiety.
In our studies reported here, by using NaI in the crystallization mixture we examined whether the cisplatin or carboplatin was converted to an iodo form with I atoms bound to the platinum centre. For our NaI co-crystallization conditions we see a PtI 2 X 2 species at His15, i.e. transiodoplatin bound to the His15 N atom. In addition, though, we also see a third platinum moiety, a PtI 3 X species, which is bound in a crevice between molecules A through residues Asn106, Arg112, Pro70, Asn65 and Asn103, and Arg128, Cys6, Glu7 and Ala10 of a symmetry-related molecule. This binding is not seen at the same residues for molecule B. The evidence that a platinum is seen bound to three I atoms is based on the presence of anomalous difference electron-density peaksand then on their refined occupancies. The fourth bound atom, i.e. the X in PtI 3 X, is interpreted as a Cl atom (cisplatin) or a portion of the CBDC moiety (carboplatin). This is reminiscent of Zeise's salt PtCl 3 C 2 H 4.
Finally, we note that in our companion paper on HEWL with Pt hexahalides we see a PtI 3 species bound to the His15 N atom.
# Conclusions
Co-crystallization of HEWL with cisplatin and carboplatin in NaI was carried out, with the XRD results showing that both platinum compounds indeed underwent a partial chemical conversion. In both cases this was to transiodoplatin (i.e. PtI 2 X 2 ) bound to the His15 N atom, with either a Cl atom (cisplatin) or a portion of the CBDC moiety (carboplatin) bound at the fourth Pt ligand atom position. These results are similar to our previous results using NaBr crystallization conditions. In the NaI conditions, only binding to the N atom of His15 is seen; this is attributed to the inability of iodide ions to produce an imidazolyl ion, at least at the pH used in these experiments.
A PtI 3 X species is seen bound in a crevice between symmetryrelated protein molecules; here the platinum is bound to three I atoms based on the presence of anomalous difference electron density and their refined occupancies, with the fourth bound atom being a Cl atom (cisplatin) or a portion of the CBDC moiety (carboplatin). The observation of these iodinated platinum species suggests the prospect of undertaking a dual K-edge (platinum and iodine) targeted radiation therapy experiment against a tumour. Cisplatin administration has previously been studied in this way involving radiation therapy using a synchrotron source and the Pt K edge alone; see also http://www.esrf.eu/ UsersAndScience/Publications/Highlights/2004/Imaging/Ima10/). To undertake such a dual K-edge radiation therapy involving iodinated cisplatin or carboplatin under patient conditions would build upon experiments such as are already under way at the ESRF with cisplatin as cited above.
# Related literature
The following reference is cited in the Supporting Information for this article:.
[table] Table 1: X-ray crystallographic data and final protein plus ligand model-refinement statistics for both the cisplatin and carboplatin crystals.Values in parentheses are for the last shell. [/table]
[table] Table 2: Occupancy values of the Pt and I atoms for both the cisplatin and carboplatin cases calculated using SHELX. [/table]
|
Successful treatment of severe COVID-19 pneumonia, a case series with simultaneous interleukin-1 and interleukin-6 blockade with 1-month follow-up
Interleukin (IL)-6 and IL-1 blockade showed beneficial results in patients with severe COVID-19 pneumonia and evidence of cytokine release at the early disease stage. Here, we report outcomes of open-label therapy with a combination of blocking IL-6 with tocilizumab 8 mg/kg up to 800 mg and IL-1 receptor antagonist anakinra 100-300 mg over 3-5 days. Thirty-one adult patients with severe COVID-19 pneumonia and signs of cytokine release, mean age 54 (30-79) years, 5 female, 26 male, and mean symptom duration 6 (3-10) days were treated. Patients with more than 10 days of symptoms, evidence of bacterial infection/ elevated procalcitonin and other severe lung diseases were excluded. Computed tomography (CT) scans of the lung were performed initially and after 1 month; inflammatory activity was assessed on a scale 0-25. Twenty-five patients survived without intubation and mechanical lung ventilation, two patients died. C-reactive protein decreased in 19/31 patients to normal ranges. The mean activity CT score decreased from 14 (8-20) to 6 (0-16, n = 16). In conclusion, most of our patients recovered fast and sustained, indicating that early interruption of cytokine release might be very effective in preventing patients from mechanical ventilation, death, and long-term damage.
# Introduction
The novel coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), was first identified in December 2019 as the cause of a respiratory illness designated coronavirus disease 2019, or COVID-19. [bib_ref] The COVID-19 pandemic: a comprehensive review of taxonomy, genetics, epidemiology, diagnosis, treatment,..., Helmy [/bib_ref] Most patients have mild to moderate disease, but some patients become severely ill. This severe disease is driven by immunoreaction and no more by virus-replication. This concept is further supported, as anti-viral treatment with remdesivir, and treatment with anti-COVID-19-spike antibodies (monoclonal or re-convalescence sera) did not improve the disease. [bib_ref] Remdesivir for the treatment of covid-19 -final report, Beigel [/bib_ref] [bib_ref] SARS-CoV-2 neutralizing antibody LY-CoV555 in outpatients with covid-19, Chen [/bib_ref] [bib_ref] REGN-COV2, a neutralizing antibody cocktail, in outpatients with covid-19, Weinreich [/bib_ref] [bib_ref] A randomized trial of convalescent plasma in covid-19 severe pneumonia, Simonovich [/bib_ref] Severe and life-threatening COVID-19 pneumonia is often characterized by local and systemic immune-mediated hyperinflammation, so-called cytokine storm. [bib_ref] On the alert for cytokine storm: immunopathology in COVID-19, Henderson [/bib_ref] [bib_ref] An inflammatory cytokine signature predicts COVID-19 severity and survival, Valle [/bib_ref] [bib_ref] Biologic therapy in COVID-19, Gonzalez-Gay [/bib_ref] [bib_ref] Tocilizumab: from the rheumatology practice to the fight against COVID-19, a virus..., Gonzalez-Gay [/bib_ref] SARS-CoV-2 is thought to bind to toll-like receptors, activates the inflammasome and the cleavage of prointerleukin (IL)-1β by caspase-1, followed by the production of active mature IL-1β, a powerful pro-inflammatory cytokine that starts an inflammatory cascade, overproduction of other proinflammatory cytokines including IL-6 and active leukocyte migration in the areas affected by inflammation. [bib_ref] Inflammasomes and its importance in viral infections, Shrivastava [/bib_ref] Higher levels of IL-6 have been positively correlated with severe course of COVID-19, whereas lower levels of IL-6 have been correlated with mild disease. [bib_ref] Elevated interleukin-6 and severe COVID-19: a metaanalysis, Aziz [/bib_ref] [bib_ref] Elevated interleukin-6 is associated with severity of COVID-19: a metaanalysis, Zhu [/bib_ref] Elevated levels of IL-6 have been found to be correlated with the likelihood of mechanical ventilation. [bib_ref] Elevated levels of IL-6 and CRP predict the need for mechanical ventilation..., Herold [/bib_ref] Activated monocytes migrating to lungs cause the Successful treatment of severe COVID-19 pneumonia, a case series with simultaneous interleukin-1 and interleukin-6 blockade with 1-month follow-up typical opac infiltrates, which lead to a reduction of oxygen uptake resulting in severe cases in respiratory insufficiency. [bib_ref] Clinical course and risk factors for mortality of adult inpatients with COVID-19..., Zhou [/bib_ref] Serum levels of biomarkers indicating the cytokine storm such as IL-6 and ferritin correlate with severe courses of the COVID-19 disease. [bib_ref] Cytokine elevation in severe and critical COVID-19: a rapid systematic review, meta-analysis,..., Leisman [/bib_ref] The hyperinflammatory nature of the lung affection in COVID-19 builds the rationale for the use of immune-modulating drugs in this disease.
Among immune-modulating drugs, which were studied in COVID-19, are corticosteroids, IL-6, and IL-1 as well as Janus kinase (JAK) inhibitors. A meta-analysis that included 1703 patients showed benefit for patients with severe acute respiratory distress syndrome when different regimes of corticosteroids were used. [bib_ref] Association between administration of systemic corticosteroids and mortality among critically ill patients..., Sterne [/bib_ref] In this prospective meta-analysis of clinical trials of critically ill patients with COVID-19, administration of systemic corticosteroids, compared with usual care or placebo, was associated with lower 28-day allcause mortality. For anti-IL-6 therapies, several trials were published with heterogeneous study populations and heterogeneous results. [bib_ref] Efficacy of tocilizumab in patients hospitalized with covid-19, Stone [/bib_ref] [bib_ref] Effect of tocilizumab vs usual care in adults hospitalized with COVID-19 and..., Hermine [/bib_ref] [bib_ref] Effect of tocilizumab vs standard care on clinical worsening in patients hospitalized..., Salvarani [/bib_ref] [bib_ref] Historically controlled comparison of glucocorticoids with or without tocilizumab versus supportive care..., Ramiro [/bib_ref] [bib_ref] Tocilizumab among patients with COVID-19 in the intensive care unit: a multicentre..., Biran [/bib_ref] [bib_ref] Tocilizumab in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled,..., Abani [/bib_ref] In the largest trial so far, with more than 2000 patients allocated to tocilizumab and more than 2000 patients allocated to standard of care (RECOVERY), tocilizumab improved survival and other clinical outcomes: 596 (29%) in the tocilizumab group and 694 (33%) in the usual care group [most of the patients (82%) in both groups received systemic steroids] died within 28 days. [bib_ref] Tocilizumab in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled,..., Abani [/bib_ref] Although the difference mentioned above has reached statistical significance, the absolute mortality risk reduction was rather low. Because of these data, the National Institute of Health is now recommending tocilizumab as a single intravenous dose in combination with dexamethasone in patients who are exhibiting rapid respiratory decompensation because of COVID-19 if not longer than 3 days hospitalized, not longer than 24 h at the intensive care unit (ICU) and who require ventilation assistance; or who were recently hospitalized but not admitted to the ICU who have rapidly increasing oxygen needs and require non-invasive ventilation or high-flow nasal cannula (HFNC) oxygen and who have significantly increased markers of inflammation [C-reactive protein (CRP) ⩾75 mg/l].
The blockade of IL-1 has been shown to be effective in several immune-mediated disorders characterized by the development of a cytokine storm [bib_ref] Benefit of anakinra in treating pediatric secondary hemophagocytic lymphohistiocytosis, Eloseily [/bib_ref] and has been reported to be successful in improving survival based in COVID-19 on retrospective cohort studies. [bib_ref] Anakinra for severe forms of COVID-19: a cohort study, Huet [/bib_ref] [bib_ref] Anakinra after treatment with corticosteroids alone or with tocilizumab in patients with..., Aomar-Millán [/bib_ref] [bib_ref] Interleukin-1 blockade with high-dose anakinra in patients with COVID-19, acute respiratory distress..., Cavalli [/bib_ref] [bib_ref] Interleukin-1 and interleukin-6 inhibition compared with standard management in patients with COVID-19..., Cavalli [/bib_ref] Furthermore, one placebo-controlled trial with 594 patients with a median disease duration of 9 days demonstrated efficacy of anakinra superior to placebo. [bib_ref] Early treatment of COVID-19 with anakinra guided by soluble urokinase plasminogen receptor..., Kyriazopoulou [/bib_ref] Therefore, Anakinra is approved for COIVDpneumonia since December 2021.
Thus, based on the pathophysiological observations in COVID-19 and evidence for the efficacy of IL-6 and IL-1 in the hyperinflammatory COVID-19 disease and based on the fact that with the single blockage of one of these cytokines results in a rather small reduction of mortality, our hypothesis was that not all immune activation pathways may be blocked in these patients with severe hyperinflammation comparable with macrophage activation syndromes. Thus, we proposed a hypothesis that a dual blockade of both key cytokines (IL-6 and IL-1) might me associated with a favorable outcome in severe COVID-19 pneumonia as compared with the single blockage of one of these cytokines.
# Methods
From May 2020 until April 2021 adult patients, aged ⩾18 years, hospitalized with moderate to severe COVID-19 pneumonia (confirmed by computed tomography -CT), duration of typical symptoms of ⩽10 days, decreased oxygen saturation (⩽95% with at least 4 l oxygen supplementation) and elevated serum markers indicating systemic hyperinflammatory reaction [ferritin, CRP, lactate dehydrogenase (LDH), and IL-6] were identified as potential candidates for a dual cytokine blockade. The presence of COVID-19 was confirmed by quantitative real-time-polymerase chain reaction (RT-PCR). We excluded patients with evidence of bacterial infection, patients with positive procalcitonin (PCT) values, patients with concurrent cardio-pulmonal disorders such as severe chronic obstructive pulmonary disease (COPD), grade III-IV and heart failure of grade II and higher according to the New York Heart Association (NYHA) classification, and patients who were more than 80 years old. All patients were treated in clinical routine and were informed about the off-label character of the therapy in reference to the statement of the Robert Koch Institute regarding therapies in 'off-label use' in COVID-19 patients after careful consideration of the individual risk-benefit ratio, and written individual consents about the off-label use were obtained. Written informed consents were also obtained according to an institutional
## Laboratory examinations
Routine laboratory investigations were performed on regular base including PCT levels, CRP, ferritin, LDH, full blood count, and IL-6.
In addition, monocytes from four patients were examined for intracellular production of IL-1β, IL-6, and tumor necrosis factor (TNF). For this, thawed peripheral blood mononuclear cells were rested overnight in complete medium at 37°C in a humidified 5% CO 2 /70% O 2 air atmosphere and incubated for additional 5 h in the presence of brefeldin A and monensin. Immunofluorescent staining of intracellular cytokines and cell markers was then performed, followed by flow cytometric analysis of gated CD68 + HLA-DR + monocytes. Data were acquired on the FACSCanto (Becton Dickinson, BD) and analyzed with FlowJo software version 10.7.1.
## Chest ct examinations
All 31 patients underwent chest CT examinations on admission and 16 patients had follow-up CT examinations 1 month after therapy onset. All CT images were reviewed by a single radiologist (JLV) with over 10 years of experience in chest imaging. Findings of COVID-19 pneumonia were assessed according to Radiological Society of North America expert consensus document. [bib_ref] Radiological society of North America expert consensus statement on reporting chest CT..., Simpson [/bib_ref] A semiquantitative scoring system was used to evaluate the pulmonary involvement of all these abnormalities on the basis of the area involved and each of the five lung lobes was visually scored on a scale of 0-5, with 0 indicating no involvement; 1, less than 5% involvement; 2, 5-25% involvement; 3, 26-49% involvement; 4, 50-75% involvement; and 5, more than 75% involvement. The total CT score was the sum of the individual lobar scores and ranged from 0 (no involvement) to 25 (maximum involvement). [bib_ref] Time course of lung changes at chest CT during recovery from coronavirus..., Pan [/bib_ref] We used the following group cutoffs to aggregate the total CT score: ⩽7 mild, 8-17 moderate, and ⩾18 severe changes.Chronic lung changes were scored subjectively into four groups (absent, mild, moderate, and severe).
Included patients received a monoclonal antibody against IL-6-receptor tocilizumab 8 mg/kg body weight with a maximum dose of 800 mg in patients with more than 100 kg body weight as a single intravenous infusion followed by an IL-1 receptor antagonist anakinra given in a dose of 100-300 mg per day for 3-5 days.
Furthermore, all patients were treated according to the actual valid guidelines, which were modified over time regarding use of remdesivir, corticosteroids, time point of admission to the ICU, and type of ventilation and position of the patients. Therefore, only one patient received additional remdesivir and 19 patients received corticosteroids.
# Results
## Clinical course
The mean age of our patients was 54 (range = 30-79) years, five female, 26 male, with a mean symptom duration of 7 (range = 3-10) days agreed to be treated so far. Mean oxygen saturation without oxygen supplementation was 83% (range = 40-95%) before initiating therapy. Mean serum ferritin level was 1417 µg/l (range = 52.4-2639), mean IL-6 was 265 ng/l (range = 2.2-2504), and CRP was 107 mg/l (range = 12.0-290.9) (see [fig_ref] Table 1: Baseline characteristics of 31 patients treated with simultaneous interleukin-1 and interleukin-6 blockade [/fig_ref].
All patients received tocilizumab as a single intravenous infusion as described above followed by anakinra 100 mg subcutaneously (sc.) over 3 days in two patients and 300 mg sc. over 5 days in all other patients. In 25 patients, we were able to prevent intubation and mechanic ventilation, 6 patients had to be mechanically ventilated, and of whom 3 patients were on extracorporeal membrane oxygenation (ECMO) ventilation management. Two of our patients died, while on ECMO ventilation. Five patients did not need to be referred to the ICU at all. The mean duration of hospital stay was 18 days (range = 5-73 days), one patient needed oxygen supplementation at the time of discharge, and one patient was referred to an external weaning center.
Laboratory results CRP levels decreased in 19/31 patients after treatment to normal ranges (<5 mg/l) and in 23/31 patients below 10 mg/l with a mean CRP of 12.9 mg/l (range = 1.8-83.1) at the end of therapy. Also, IL-6 levels, ferritin, and LDH dropped in all patients after treatment [fig_ref] Table 2: Course of laboratory parameters and chest CT changes in 31 patients with... [/fig_ref].
After 1 month, CRP levels were measured in 19/31 patients and normalized in 14 of these patients with a mean CRP of 9.6 mg/l (0.3-74.0). Mean IL-6 levels were 16.9 ng/l (1.5-74), mean LDH was 297.7 U/l (155-501), and mean ferritin was 408.4 µg/l (54-1872) [fig_ref] Table 2: Course of laboratory parameters and chest CT changes in 31 patients with... [/fig_ref].
Intracellular cytokine production over time was determined in monocytes from four patients. Levels of IL-6 and IL-1β but not that of TNF decreased during treatment .
## Chest ct examinations
The mean severity score for active COVID-19 lung changes in the acquired chest CT scans was 15 (range = 7-23, n = 31) at baseline and decreased from 14 (range = 8-20, n = 16) to 6 (range = 0-16, n = 16) 1 month after treatment. Chronic fibrotic changes were absent in any patient at baseline and mild in six patients 1 month after intervention and [fig_ref] Table 2: Course of laboratory parameters and chest CT changes in 31 patients with... [/fig_ref].
## Safety
In 21 of 31 patients, no adverse events were noticed, and 2 patients died on ECMO because of bacterial infection and bleeding. The following adverse events were seen in the remaining eight patients: lower abdominal pain, urinary tract infection, and gastrointestinal bleeding because of antrum ulcers Forrest III in one patient, atrial fibrillation, urinary tract infection, and apoplexy in two patients, and in the other six patients, bacterial infections occurred: increased PCT and antibiotic treatment, bacterial pneumonia, and pleural empyema, one with additional fungal lung infection with candida.
# Discussion
In this case series, most of our patients treated with both tocilizumab and anakinra recovered mostly fast and sustained. Unfortunately, two patients died while they were treated on ECMO device, one because of bacterial infection and another one because of lung bleeding. Further four cases temporarily needed mechanical ventilation after treatment initiation. This is clearly superior when compared with the large trials with tocilizumab (RemapCap and Recovery). [bib_ref] Tocilizumab in patients admitted to hospital with COVID-19 (RECOVERY): a randomised, controlled,..., Abani [/bib_ref] In the RemapCap trial, with a quite similar selection of patients, the in-hospital mortality for the tocilizumab treated group was 27% and for the control group 36%, and in the Recovery trial, it was 29% in the tocilizumab group and 33% in the control group. Of note, in both trials more than 80% were additionally treated with corticosteroids. In our case series, only 61% patients were treated with dexamethasone. This was because the guideline for the use of dexamethasone was introduced later in 2020; some of our patients were already treated before.
We selected only patients with short disease duration of ⩽10 days and with evidence of respiratory insufficiency and signs of cytokine storm. Patients with signs of bacterial super-infection (PCT >0.5) were excluded. We hypothesize that early initiation of treatment in patients presenting with signs of hyperinflammation is crucial to disrupt the inflammatory cascade and to prevent severe lung The studies with tocilizumab and anakinra in COVID-19 as single substance were conducted without restriction of the disease duration that most likely was the reason for the heterogeneous study outcomes. [bib_ref] Tocilizumab among patients with COVID-19 in the intensive care unit: a multicentre..., Biran [/bib_ref] [bib_ref] COVACTA trial raises questions about tocilizumab's benefit in COVID-19, Furlow [/bib_ref] In the COVACTA trial, the mean days form symptom onset to baseline were 12.1-11.4 days.In one recent trial, patients were included who were treated with tocilizumab, sarilumab, or placebo in addition to steroids within hours after admission to the ICU, which occurred only 1.2 (0.9-2.8) days after being hospitalized. In this study, the hospital mortality was clearly reduced to 28.0% (98/350) for tocilizumab and to 22.2% (10/45) for sarilumab when compared with 35.8% (142/397) in the control group.
In the SAVE MORE trial, 34 the 28 days mortality rate was low in both the placebo (6.9%) and the anakinra arm (3.2%) maybe because patients were treated early and were clearly identified to have active disease with soluble urokinase plasminogen activator receptor (suPAR) stratification.
Remarkably, even with IL-6 blockade given on the background of usual care (nowadays, normally including steroids), the mortality in severe COVID-19 disease remains with almost onethird of the patients very high. In our uncontrolled observation, two of the patients (6.4%) died. With all the limitations related to the small numbers and lack of control, the dual cytokine blockade represents a promising approach deserving a test in a randomized controlled trial.
Therapy of simultaneous use of tocilizumab and anakinra was safe. In 21 patients, no side effects were observed. In the 10 patients with adverse events, apoplexy is a typical manifestation of the COVID-19 disease and all but one of infections recovered, indicating that the combination of these two drugs was safe and well tolerated even in combination with corticosteroids in our patients. Therapy on ECMO is often fatal, which was also the case in our two patients who unfortunately progressed to ECMO. These two patients were male, suffered from co-morbidities, especially from severe adipositas.
The applied therapy was associated with reduction of systemic markers of inflammation as well as of inflammation detected by CT. This could be also measured by immediate decrease intracellular of IL-6 and IL-1β cytokine production in monocytes in four patients. The chest CT severity score was reduced by more than 60% 1 month after treatment. Also a subjective score for chronic pulmonary changes indicated that nearly no chronic changes occurred in the treated patients after 1 month, only in 6 of 16 patients mild fibrotic changes were found.
The median 12 days of stay in the hospital in our group was clearly less than what was published for a similar group of patients with severe course of the disease but without treatment in Germany, which was 18.1 days for patients with non-invasive ventilation [bib_ref] Case characteristics, resource use, and outcomes of 10 021 patients with COVID-19..., Karagiannidis [/bib_ref] and less than in the COVACTA Study in which the median time to hospital discharge was 20-28 days in the survivors.In our group, patients experienced a longer hospitalization because of stroke, bacterial infection, and because of rehabilitation.
Our case series was rather small, and we treated off label. The strength of the case series was that we treated consecutively a homogeneous, wellcharacterized group of patients with moderate to severe COVID-19 pneumonia. Ín 15 patients, we also had CT follow-up investigations as another objective parameter of therapeutic success of our treatment approach.
In conclusion, therapy with the combination of tocilizumab and anakinra in patients with severe course of COVID-19 presenting with a hyperinflammatory syndrome resulted in fast and sustained improvement of signs, symptoms, and markers of inflammations and prevented most of these patients from fibrotic changes in the lung as seen by CT with a favorable safety profile. Larger controlled trials should be conducted to confirm the role of dual cytokine blockade in severe COVID-19 disease.
# Declarations
Ethics approval and consent to participate All patients were treated in clinical routine and were informed about the off-label character of the therapy in reference to the statement of the Robert Koch Institute regarding therapies in 'off-label use' in COVID-19 patients after careful consideration of the individual risk-benefit ratio, and written individual consents about the off label use were obtained. Written informed consents were also obtained according to an institutional protocol of the Charité (Ethical Committee of Charité Universitätsmedizin Berlin approval no. EA2/072/29). Patients' information was de-identified for demographic and other patient information. In addition, the equator guideline for case series used (CARE) was followed, and the checklist is added as a Supplementary Material. [bib_ref] Early treatment of COVID-19 with anakinra guided by soluble urokinase plasminogen receptor..., Kyriazopoulou [/bib_ref]
## Consent for publication
The patients consent included the consent for publication.
# Author contributions
[fig] Figure 1 7, Figure 2: Monocyte production of interleukin (IL)-6, IL-1β, and tumor necrosis factor-alpha (TNF-α) as well as serum levels of IL-6 and ferritin over time in four patients before and during treatment with simultaneous tocilizumab and anakinra. Serum levels are given relative to the upper limit of normal ranges.MFI, mean fluorescence intensity; se, serum. H Haibel, D Poddubnyy et al. journals.sagepub.com/home/tab Chest CT of three patients with active COVID-19 infection before (a) and 1 month after treatment (b) with a combination of tocilizumab and anakinra. [/fig]
[table] Table 1: Baseline characteristics of 31 patients treated with simultaneous interleukin-1 and interleukin-6 blockade. [/table]
[table] Table 2: Course of laboratory parameters and chest CT changes in 31 patients with severe COVID-19 pneumonia and therapy with simultaneous interleukin-1 and interleukin-6 blockade. [/table]
|
Falling living standards during the COVID-19 crisis: Quantitative evidence from nine developing countries
Colombia). We document declines in employment and income in all settings beginning March 2020. The share of households experiencing an income drop ranges from 8 to 87% (median, 68%). Household coping strategies and government assistance were insufficient to sustain precrisis living standards, resulting in widespread food insecurity and dire economic conditions even 3 months into the crisis. We discuss promising policy responses and speculate about the risk of persistent adverse effects, especially among children and other vulnerable groups.
# Introduction
As a result of the COVID-19 pandemic, economic activity has contracted around the globe. Fear of the virus and strict social distancing policies have led individuals in virtually all countries to modify their consumption and working habits. Economically vulnerable citizens of low-and middle-income countries (LMICs), where the majority of the world's population resides, potentially face stark threats to their livelihoods. We use survey data systematically collected from 16 samples of over 30,000 households (containing more than 100,000 people) in nine countries in Africa, Asia, and Latin America to provide a rich, quantitative description of the economic effects of COVID-19 among various subpopulations in these LMICs.
There is reason to believe that rich and poor countries are experiencing the crisis very differently, which makes systematic documentation of the effects of COVID-19 in distinct settings critical. In industrialized nations, economic losses are often mitigated by government protection programs, employer adjustments to hours or compensation, or household savings. Absent broad social safety nets, declines in economic activity in LMICs can have more adverse welfare consequences, especially for those working in the informal economy. On the other hand, epidemiological models predict that health impacts of the virus may be weaker in LMICs given their relatively youthful populations. Poorer countries are also generally less connected to the global economy through trade and travel and, thus, were exposed to the pandemic later with valuable time to prepare and learn from the experiences of China, Europe, and North America (although it is unclear whether these opportunities were seized in practice). cite these factors to argue, using macroeconomic and financial statistics, that the initial economic effects of the pandemic were unexpectedly mild in LMICs. In contrast, international organizations have used similar aggregate data to make dire projections about gross domestic product (GDP) losses, decreases in remittance flows, and increases in poverty and hunger. Yet aggregate data have recognized deficiencies relative to direct surveying for tracking the well-being of the poor. In this study, we therefore rely on original, large-sample, representative household surveys to assemble a systematic and in-depth look at how the pandemic affected people's lives in LMICs in the months following the global outbreak.
Household surveys are necessary because aggregate data can overlook large segments of the population. Over a quarter of economic activity and half of all workers in Africa, Asia, and Latin America are in the informal sector, and therefore are not captured in most official statistics. Informality similarly undermines the informativeness of measures of private sector transactions such as payroll, credit, or smartphone transfers. Hence, the approaches national statistical agencies and researchers [e.g.,] have used to document the economic losses from COVID-19 in industrialized nations cannot easily be implemented in LMICs. Many LMICs rely on periodic household or labor force surveys to measure economic activity. However, the low frequency of these surveys makes them insufficient for real-time tracking during a crisis.
Our research team rapidly adapted existing data collection protocols to deploy phone surveys starting in early April 2020 to track economic outcomes during the COVID-19 crisis. We use random sampling to generate statistically representative information about 2 of 12populations in nine countries in Africa, Asia, and Latin America. The surveys cover heterogeneous samples constructed in different ways. Seven samples rely on random phone digit dialing (RDD) and skew toward wealthier and more educated mobile phone owners, while the other nine are drawn from earlier studies representative of specific subsamples, including formal and informal sector workers, agricultural laborers, small business enterprises, refugees, migrants, and their families. While the magnitude of effects varies across settings, the data reveal a consistent picture of heightened economic distress that spans both geography and socioeconomic strata.
The combined dataset documents steep drops in employment and income in our samples that rival or exceed economic losses experienced in the United States and other rich nations [e.g.,]. A full 50 to 80% of sample populations in Bangladesh, Burkina Faso, Colombia, Ghana, Kenya, Rwanda, and Sierra Leone report income losses during the COVID-19 period. If these effects persist, then they risk pushing tens of millions of already vulnerable households into poverty. This economic shock caused by the global pandemic spans socioeconomic strata within each country, and similar proportions of households at different rungs of the socioeconomic ladder report employment and income decreases. While the data shed light on the proportion of population suffering an income decline during the period, it should be noted that we are only able to characterize the magnitude of the decline for specific households in some of the samples.
By April, many households were already unable to meet basic nutritional needs. For example, 48% of rural Kenyan households, 69% of landless agricultural households in Bangladesh, and 87% of rural households in Sierra Leone were forced to miss meals or reduce portion sizes to cope with the crisis. Comparing to preexisting baseline data verifies that these levels greatly exceed the food insecurity normally experienced at this time of year. If anything, COVID-19 fortunately hit during a "postharvest" period in South Asia, when many people have grain stocks to draw down. High levels of food insecurity may continue to worsen as the crisis persists through the agricultural cycle.
The data paint a consistent picture: The economic shock and attendant disruptions to livelihoods during the early stages of pandemic appear to be large across a range of populations in Africa, Asia, and Latin America. The scale of the disruption may even exceed the effects that economists have documented in other recent global crises, including the 1997 Asian Financial Crisis, the 2008 Great Recession, and the Ebola outbreak of 2014.
As a result, for LMICs, the economic crisis precipitated by COVID-19 may become as much a public health and societal disaster as the pandemic itself. The link from severe economic crisis during childhood to subsequent deterioration in adult health, nutrition, education, and earnings capacity has been demonstrated in many contexts. Almonddocuments notable declines in education and adult earnings among those in utero during the 1918 influenza pandemic, and Maccini and Yangshow that children born during periods of weather-related economic hardship in Indonesia experience worse health, educational achievement, and income as adults. A growing body of long-run experimental research specifically links childhood nutrition to standards of living during adulthood [see]. These channels of long-run transmission indicate that without mitigation, the substantial and widespread economic distress caused by the current pandemic may induce fallout that persists for decades into the future.
# Data and methods
## Sample construction
We present results from 16 samples in nine LMICs, as enumerated in. These countries-Bangladesh, Burkina Faso, Colombia, Ghana, Kenya, Nepal, Philippines, Rwanda, and Sierra Leone-have a combined population of nearly 500 million. We study several different subpopulations in Bangladesh and Kenya, and half of our study samples are drawn from two countries with a combined population of 200 million. In each sample, we conducted at least one telephone survey sometime during the period April-June 2020, after the outbreak of COVID-19 and the initial implementation of government lockdowns or other social distancing policies. Appendix A in the Supplementary Materials provides information on the timing of the spread of COVID-19 and government-imposed mobility restrictions relative to survey implementation dates across all countries in our sample.
Nine of the samples (denoted BGD1-5, KEN1, KEN2, NPL1, and SLE1 inin four countries were constructed from preexisting studies and were randomly drawn to be statistically representative of the population of interest for those earlier studies. An advantage of this is that we have additional pre-COVID-19 baseline data on living standards for these nine samples. When possible, survey questions were designed to be comparable to preexisting baseline data. In two of these samples (BGD5 and NPL1), we can compare the time path of outcomes to the typical seasonal pattern observed in prior years. The populations from which these samples are drawn vary, as described inand in more detail in Appendix A in the Supplementary Materials.
Six other samples (BFA1, COL1, GHA1, PHL1, RWA1, and SLE2) are drawn via phone RDD, making them statistically representative of the set of active mobile phone numbers held by adults. Household surveys of mobile phone usage in various countries shows that the vast majority of adults in the sample countries and in LMICs more broadly now have access to mobile phones. For these samples, baseline data are constructed from respondent recall anchored to the introduction of substantial and memorable policy restrictions, typically the closure of schools. The final sample (KEN3) is a hybrid of 4052 households sampled from those in the 2015/16 Kenya Integrated Household Budget Survey who provided telephone numbers and an additional 767 adults contacted via RDD.
The primary reporting unit varies by sample. In samples constructed from existing study populations, household membership was already known, and the household is the unit of focus. A disadvantage is that this does not allow us to disaggregate all effects by gender, but some surveys collect outcomes specific to women and children (e.g., on domestic violence) that we will report below. In the RDD-based samples, data represent the individual adult associated with the phone number, with some limited questions about that individual's household. Sampling weights for representativeness are described in Appendix A in the Supplementary Materials.
## Survey methods and timing
All post-COVID-19 data were collected via telephone interviews to minimize in-person contact and comply with government social distancing guidelines. Interviews were conducted by local enumerators in each country, with procedures to match languages, dialects, and accents between respondent and enumerator. Surveying by phone made rapid and large-scale data collection possible over large geographical units. In two samples (KEN1 and SLE1), we conduct high-frequency surveys spanning a long enough period to examine the evolution of post-COVID effects over time.
Unfortunately, interviewing by telephone places limits on data collection. Surveys were designed to be short, lasting only 15 to 30 min with relatively coarse measures of income and welfare, and render anthropometric measurements infeasible. Moreover, very poor households, who may not own phones or live in areas with low connectivity, may be underrepresented. In the more agriculture-dependent nations in our sample, the baseline pre-COVID-19 period falls during a postharvest season in Bangladesh, Burkina Faso, Ghana, Nepal, and Sierra Leone, when food stocks in the rural population tend to be relatively high, and food prices are typically low. While this timing is fortunate for households' ability to cope, postharvest is also a low point for agricultural labor demand in these countries, making subsequent declines in income and employment particularly notable. In Kenya, many farm households were already entering their "lean" season at the start of the crisis in March 2020. Natural seasonal variation in outcomes can complicate empirical inference on the true effects of COVID from these data, an issue we will address in greater depth below.
## Construction of outcome variables
Survey questions were coordinated across samples to the extent possible for data consistency. However, there were necessary adaptations for local context, and some questions were altered to conform to preexisting baseline data. To maximize consistency in reporting outcomes, we present each main result as the fraction of respondents reporting a change in an outcome post-COVID-19 relative to the preperiod. We provide a detailed description of how each variable was constructed for each sample in Appendix C in the Supplementary Materials and discuss the robustness of results to other reasonable ways of defining outcomes in Appendix B in the Supplementary Materials.
To examine heterogeneity of effects across socioeconomic strata, we further subdivide our analysis by socioeconomic status (SES) within sample. This classification is based on the within-sample median pre-COVID-19 consumption expenditure for six samples (BGD1-4, KEN1, and SLE1) and median pre-COVID-19 income for four samples (BGD5, KEN2-3, and NPL1). For the remaining six samples, high and low SES are distinguished by respondents' scores on a "Poverty Probability Index" derived from the most recent national household survey, calibrated to either 100 or 200% of the national poverty line.
We calculate a "drop in income" measure at the household level in half of the samples (BGD1, BGD4, BGD5, KEN1-3, NPL1, and SLE1) and at the individual level in the remaining half (BGD2, BGD3, BFA1, COL1, GHA1, PHL1, RWA1, and SLE2). Four of the surveys (BGD2, BGD3, BGD5, and NPL1) compare income reported during pre-COVID-19 baseline surveys to income reported during the post-COVID-19 telephone interview to determine whether there has been a drop in income. The other surveys use retrospective reports of baseline income collected during the post-COVID-19 survey to compare with current income. We note that these retrospective reports carry the risk of respondent recall or reporting biases, which we would expect might lead to an overestimate of the extent of declines in income in some cases, a possibility we discuss further in the "Cross-survey comparability, representativeness, and limitations of the data" section.
We construct a "drop in employment" measure at the individual level in nine surveys (BGD2, BGD3, BFA1, COL1, GHA1, PHL1, RWA1, SLE1, and SLE2), with a drop defined as a respondent who reported working during the pre-COVID-19 reference period but not working at the time of the post-COVID-19 interview. Five surveys (BGD1, BGD4, and KEN1-3) measure the change in employment at the household level. In BGD1 and BGD4, a drop in employment is registered if any adult in the household was working during the baseline period but no adult was working during the post-COVID-19 period. In KEN1, a drop in employment is recorded if any adult in the household reports losing their job since February 2020 and is not currently working. The individual measures of employment declines are more strict than the household measures (with the exception of BGD1 and BGD4).
"Reduced access to markets" is measured at the household level based on respondent reports that they or any household member faced difficulties in purchasing food because of mobility restrictions, closed markets, or food shortages. Measures of food insecurity that we label "missed or reduced meals," available for all samples, are based on respondents' reports that they or someone in their households skipped meals or reduced portion size or quality. The reference period is the past 7 days unless otherwise noted in the appendix. In five surveys, households were classified as missing or reducing meals only if the respondent reported being unable to buy essential food items because of a lack of resources over the past 7 days (for BGD1-4) or 14 days (for NPL1). In the other surveys, there is no restriction on the reason for reducing or skipping meals.
Measures of "receipt of government or NGO support" are based on reports that the respondent (BFA1, COL1, GHA1, PHL1, RWA1, and SLE2) or the household (BGD1-5 and KEN1-3) have received food, cash, or other support from government programs or NGOs over the past month (BFA1, COL1, GHA1, PHL1, RWA1, and SLE1) or 2 weeks (KEN1-3). For BGD1-3, only households who reported being unable to buy essential food items over the past week because they lacked sufficient resources were asked whether they received assistance from government or NGOs.reports this number as a share of the total sample, so is a lower bound estimate of the share of households that received such assistance in the case of BGD1-3.
Enterprise profit and revenue data in KEN1 are drawn from a parallel survey of a sample of enterprises drawn from a baseline pre-COVID-19 census of all enterprises in the villages where sample households reside. Current profits and revenues of enterprises are asked for the past 14 days; pre-COVID-19 profits and revenues are asked retrospectively for a "typical 2-week period in February 2020." Enterprise profits in SLE1 are based on the household head's selfemployment profits over the past 7 days (post-COVID-19) or over a typical week in the month before the first lockdown.
Total consumption expenditure in KEN1 is based on the value of total household food consumption over the past 7 days and nonfood expenditure over the past 14 days. Consumption expenditure in SLE1 is based on household expenditure on five main staple food items over the past 7 days. To construct the pre-COVID-19 benchmark, the same expenditure question is asked retrospectively for a period about the same time, but a year prior.
To assess effects on consumer prices, we construct a price index for the KEN1 sample from respondent reports of the prices of 20 common consumer items. In the SLE1 sample, the price index is constructed from respondent reports on the prices of the five staples used to estimate consumption.
We also collected respondent reports of domestic violence in the KEN1 sample, in cases when female enumerators interviewed female respondents. After screening questions concerning privacy, safety, and willingness to discuss sensitive issues, respondents were asked three questions regarding threats of harm, physical abuse, and sexual activity being forced by any of their partners. Domestic violence against children is based on the respondent's report that she or herdrop in employment, (3) reduced access to markets, (4) having health care access delayed, (5) having to reduce or miss the amount of meals, and (6) receiving NGO or government support. Column (7) shows the total number of households surveyed in each sample. Column (7) shows the maximum number of observations available for analysis in each study though specific measures are sometimes based on smaller samples. The division of respondents in each sample into "higher" and "lower" socioeconomic status is based on that respondent's status within each sample, based on baseline consumption expenditure (BGD1-4, KEN1, and SLE1), baseline household income (BGD5, KEN2-3, and NPL1), and a Poverty Probability Index (others). Data from KEN1 are restricted to the first round of surveys. Blank cells denote that no data were available. These results are reproduced with standard errors in Appendix in the.
## Share of households experiencing
## Cross-survey comparability, representativeness, and limitations of the data
The evidence is drawn from surveys of over 30,000 respondents in nine countries, with some coordination in questionnaire design. Yet the realities and constraints of survey work during the pandemic imply that none of the samples are nationally representative. In the seven RDD-based samples, the analysis is weighted to make the reported statistics representative of the active mobile phone numbers used by adults. While this is an increasingly broad portion of the adult population in these countries, it excludes people without access to mobile phones and may overrepresent people with multiple numbers. As a consequence, we see in a direct comparison of our survey respondents to nationally representative samples presented in Appendix table S1 in the Supplementary Materials that all of the RDD samples, with the exception of the Philippines, are much better educated than the populations of the countries from which they are drawn. In Burkina Faso, Philippines, Rwanda, and Sierra Leone the RDD samples are less poor, and in Burkina Faso, Ghana, and Sierra Leone they are more urban than the national population. Household sizes are larger in the RDD samples than in the national population in Ghana, Kenya, Philippines, Rwanda, and Sierra Leone. In the Philippines, RDD respondents are disproportionately female, while in Ghana they are more likely male. Nine of the samples are based on earlier studies and are representative of specific subpopulations within a country. Pre-COVID-19 data from these studies provide an important baseline for our analysis. These samples are not intended to be representative of the entire country, as highlighted in Appendix table S1 in the Supplementary Materials. In Bangladesh, the populations sampled in our studies are very diverse. The Rohingya refugees (BGD2) and landless laborer (BGD5) populations are much poorer than the national average, while the communities living near the refugee camps (BGD3) are relatively better off than the national population. The other Bangladesh samples (BGD1 and BGD4) are both more rural and male than the population as a whole but have average incomes near the national average. The BGD1 rural sample has much lower education on average than the national population, while the BGD4 sample of applicants for agricultural work permits in Malaysia is better educated.
The Kenya rural household and refugee samples (KEN1 and KEN2), like the populations from which they are drawn, are both more rural and poorer than the Kenya national average. The rural KEN1 sample has much lower levels of education than the national average, and while the population of refugees also has lower levels of education, selection into the phone sample is such that the secondary school completion rate of the KEN2 refugee sample is similar to the national average. KEN3 households, in contrast, are more urban and better educated than the national average, but still poorer. Note that our Kenya samples are mostly located in regions that were unaffected by the 2019-2020 locust plague (23); when we exclude from our analysis the counties that were most affected by this plague, the results remain virtually the same (results available upon request). The agricultural households surveyed in Nepal (NPL1) are all rural and much poorer than the national average. However, their secondary school completion rate is at the national average. The Sierra Leone Rural Electrification survey (SLE1) is entirely rural, consists of larger households than the national average, and has relatively few female respondents. The diversity of the samples we have gathered together makes comparisons across countries of the results more difficult. However, that same diversity provides some valuable insights into the impact of the COVID-19 pandemic across widely varying contexts.
In Appendix A in the Supplementary Materials, we characterize the timing of the first COVID-19 case and post-COVID survey dates in each country relative to the timing of the lean season. The post-COVID-19 reporting period in Burkina Faso, Kenya, Sierra Leone, and, to a lesser extent, Ghana occurs during the beginning of lean agricultural seasons. Some of the reported declines in employment, income, or food security that we observe may therefore reflect expected seasonal changes, rather than the effects of the epidemic. Current estimates of the extent of consumption seasonality in African contexts are that it is relatively modest. For example, (24) estimates a 2 to 3% decline in food consumption between the highest and lowest consumption months in Tanzania. There is, however, evidence that the quality and variety of food consumed vary more markedly across seasons than does the quantity of food. Our measure of food security focuses on quantities consumed (meals skipped or portion sizes reduced) rather than on quality and should therefore be less affected by seasonal changes.
As described more fully in Appendix B in the Supplementary Materials, some of the measures depend on respondent recall. For example, the "drop in income" variable collected in the seven RDD samples compares earnings over the past 7 days to pay in "a typical week" before government closed schools (or other marker of the onset of the COVID-19 crisis). These retrospective reports carry the risk of respondent recall or reporting biases based on their perception of the pandemic as a crisis, which could lead to an overestimate of the extent of declines in income.
In three of our samples (NPL1, BGD4, and BGD5), we can examine the extent of recall bias by comparing pre-COVID-19 outcomes as measured by survey responses elicited before versus after the onset of COVID-19. In Appendix B in the Supplementary Materials we report the three outcomes for which we have comparable datapre-COVID-19 seasonal food security in the NPL1 sample, monthly household earnings in April in the BGD4 sample, and food security in January and February in BGD5-measured in surveys from both before and after the onset of COVID-19. In all three cases, the data about pre-COVID-19 outcomes as reported in our post-COVID-19 phone surveys closely track their counterparts as reported in pre-COVID-19 surveys. This consistency gives us some confidence that the pandemic itself did not have a large influence on recall-based reporting of prior conditions. However, this is still a difficult concern to address in general for all our data, and the possibility of these biases should be kept in mind.
# Results
## Livelihoods during the covid-19 crisis
Results indocument the widespread nature of economic hardships and the decline in living standards across the nine LMICs in the study. Across the 16 samples, between 8 and 87% of respondents report a drop in income during the crisis period, with a staggering median of 70% (column 1). The proportions reporting declines in employment are similarly high, ranging from 5 to 49% with a median share of 30% (column 2). The estimated magnitude of the economic shock remains stable whether comparing to preexisting baseline data or to respondent recall about their pre-COVID status as reported to us in a phone interview conducted after COVID hit. These measures capture the share of individuals or households that experienced a drop in well-being during the pandemic period rather than the net changes in income or employment. However, the proportion of respondents reporting declines in income (median 70%) exceeds those reporting rising income during the period by an order of magnitude (median across samples 7%). Appendix B in the Supplementary Materials discusses robustness of the estimates in detail.
The adverse economic shock experienced by individuals surveyed in these countries has been compounded by impediments to livelihood. In most countries, a large share of respondents report reduced access to markets, with the median share being 31% (range, 3 to 77%; column 3), likely related to the ubiquitous lockdowns and other mobility restriction policies adopted during March through June 2020. Where data are available, meaningful shares of respondents also report delays or other difficulties accessing health care (median, 13%; column 4).
Together, these drops in employment, income, and access to markets and services appear to contribute to higher levels of food insecurity. During the survey period, between 9 and 87% of respondents were forced to miss or reduce meals (median share, 45%; column 5), an issue we examine further in the next subsection. Even in Colombia (sample COL1), the country in our sample with the highest per capita GDP and thus potentially the greatest financial resources to cope with the crisis, the majority of respondents report drops in income (87%) and employment (49%), and an increase in food insecurity (59%).
Social support in response to the economic shock has been mixed in our populations of study. Across samples, the proportion of respondents who report benefiting from government or NGO crisis support runs the gamut from 0 to 49%, with a median of 11%. However, the high rate of missed meals and reduced portion sizes suggests that even when these efforts are present, they have been insufficient. For instance, Rohingya refugees in Bangladesh (BGD2) report the highest rates of assistance, given the preexisting international aid infrastructure serving those communities. Even in this sample, 27% of respondents report food insecurity. More detailed data in one sample (KEN1) indicate that households also engage in extensive dissaving, such as selling assets and spending stored cash, to stabilize consumption.
These adverse effects on employment, income, market access, and food security vary substantially both across countries and across different subsamples within countries. For example, in the subset of national surveys, the share of households experiencing a drop in income varies across countries from 25% in Kenya to 87% in Colombia. Within the Kenya samples, the share of households experiencing drops in income ranges from 8 to 69%. Thus, the median impacts shroud significant variation across settings. Especially within countries, it is likely that this heterogeneity results, at least in part, from differences across the subgroups surveyed.
At the same time, however, we find little evidence that this variation is systematic, e.g., by socioeconomic or refugee status. In most of cases, we cannot reject equality in the share of high and low SES households affected. However, the impact of an equivalent income drop may be greater among low SES households, as evidenced by the generally higher rates of food insecurity reported in these subsamples. There is similarly no clear pattern across refugee and nonrefugee populations. Levels of reported food insecurity are actually slightly lower among refugees than the host communities living near Rohingya camps in Bangladesh (BGD2 to 3). On the other hand, food insecurity is somewhat higher among refugees in Kenya compared with a national sample (KEN2-3). More detailed data collected in BGD2-3 surveys suggest that the presence of international humanitarian organizations in the Rohingya camp areas may have helped buffer the economic shock for refugees.
## Impact timing, magnitude, and seasonality
We next describe the magnitude and timing of the effects on economic outcomes drawing on a subset of samples that feature more detailed panel or repeated cross-sectional data with richer measures of several key outcomes. Firm operations, a natural measure of overall local economic activity, appear to have been very adversely affected during the COVID-19 crisis where we have these data. In rural Kenya (KEN1), average firm profits and revenues dried up, falling by 51 and 44%, respectively (both with P < 0.05 relative to precrisis levels;. The analogous decline in Sierra Leone rural towns (SLE1) is a massive 50% (P < 0.05 relative to precrisis levels; A2). This evidence complements numbers on the share of the population experiencing any decline in employment or income inby quantifying the depth of the economic decline.
In the rural Kenya sample, there is also a pronounced decline in per capita consumption expenditures during the crisis (B1), with declines in nonfood expenditures of 29% (P < 0.05 relative to the first observation period) persisting through all of April and May 2020. During the same period, food expenditures in Kenya and Sierra Leone actually rose slightly, by 11% (B1) and 6% (B2), respectively, although in Sierra Leone, this appears to have been driven by higher food prices facing these households (19%, P < 0.05 relative to the preperiod; C2) rather than greater quantities consumed. In contrast, Kenyan prices were largely stable or even fell slightly during the same period (C1). These data indicate that households appear to be cutting back nonfood consumption in an effort to maintain essential food intake.
Examining food insecurity in greater detail, we observe rising rates of missed meals and reduced portions during the crisis in both Kenya (D1) and Sierra Leone (D2), respectively. In Kenya, we record a 38% proportional increase in the rate of adults missing meals (0.5 meals per week) and 69% for children (0.5 meals per week). The proportional increase in the share of adults reducing portions in Sierra Leone is 86% (30 percentage points) and for children is 68% (17 percentage points, P < 0.05 for all of these effects). The sharp rise in food insecurity among children is particularly alarming given the potentially large negative long-run effects of undernutrition on later life outcomes.
The crisis period has been damaging for other dimensions of child development beyond nutrition. Schools in all sample countries have been closed during most or all of the study period. Nontrivial shares of respondents report reduced access to health facilities, including prenatal clinics and vaccinations, column 4). The combination of a lengthy period of undernutrition, closed schools, and limited health care may be particularly damaging in the long run for children from poorer households who do not have alternative resources to make these critical human capital investments.
The rate of dissaving indicates there may be a range of other foregone household investments, from improved agricultural inputs to new small business opportunities. Lack of investment in both human and physical capital during a time of crisis can transmit the economic fallout of the pandemic far into the future.
The COVID-19 crisis could also have contributed to rising rates of domestic violence in the rural Kenya sample for which we have detailed survey reports (E1). Both violence against women and children-groups that are already marginalized in rural Kenyan society-rise by 4 and 13% (0.3 and 2.6 percentage points), respectively, during the crisis period, although these increases are not statistically significant. This increase in violence could generate additional negative and persistent effects on physical and mental health.
A central methodological concern in interpreting the patterns described inandis that factors other than the COVID-19 crisis could drive the evolution of outcomes over time. A leading possibility is that month-to-month seasonality, related for instance to the agricultural crop cycle, can also produce large changes over a span of a few months. It is challenging to fully address these concerns given distinct growing cycles for different crops in different countries, and sometimes even divergent harvest timings for different crops across regions within the same country. However, the consistency in outcomes across 16 different samples in nine countries on multiple continents, with a wide range of seasonal harvest and weather (and other) patterns, strongly suggests we are documenting the effects of a crisis that go beyond natural seasonal variation.
In two specific cases, we can directly contrast the excessive food insecurity experienced during the 2020 COVID-19 crisis to the natural seasonal patterns observed during those same months in previous years. In the BGD5 and NPL1 agrarian samples, there is monthly information on food insecurity during the 2016-2019 period that provides an ideal benchmark.clearly shows the pronounced seasonal variation in food insecurity in both Bangladeshand Nepalthat spikes during preharvest lean or "hungry" seasons even during "regular" years. It is also apparent that levels of food insecurity are far higher during the 2020 crisis than they were during the same season in previous years: The rate of food insecurity in Bangladesh in April 2020 is roughly twice as high as in previous years, and this season-adjusted difference is statistically significant (P < 0.05). In these cases, the leading explanation for the effects we document in 2020 is the COVID-19 crisis rather than seasonal fluctuations. It is notable that in both countries, the COVID crisis occurred during the favorable postharvest period with its relatively low level of food insecurity during normal years. Baseline levels of deprivation typically rise sharply in the final 4 months of the calendar year.
## Conclusions and implications for policy
We document pronounced declines in employment, income, and food security since April 2020 across 16 survey samples in nine LMICs, with surveys covering over 30,000 households. This study provides some of the most systematic data to date on how the outbreak of COVID-19 has affected households across multiple LMICs The post-COVID-19 levels are from questions that recall data from the prior 7 days (A to D2 and C to D1), prior 2 weeks, and a combination (prior 7 days for food and prior 2 weeks for nonfood expenditures in B1). The weeks on the horizontal axis refer to the start of the recall period for each observation rather than the period during which the data were collected. The dotted lines in A1 and A2 show the linear trend from the pre-COVID baseline to the first observation for each respective time series. Baseline level for D1 is 1.3 days out of seven for adults and 0.72 for children. Baseline level for D2 is 35% of adults missing any meals in prior 7 days and 25% of children. Baseline level for E1 is 8% of adults experiencing violence in the prior 7 days and 20% of children. *P < 0.05.
in several major world regions. While the realities of rapidly deploying a survey over the phone during a pandemic make it difficult to reach a truly nationally representative sample, we study heterogeneous samples spanning three continents. We find that the economic shock in these countries-where most people depend on casual labor to earn enough to feed their families-leads to deprivations that seem likely to generate excess future morbidity, mortality, and other adverse longer-term consequences.
The findings highlight the importance of generating on-the-ground survey data to track well-being during the crisis to gather detail necessary to craft evidence-based policy responses. We demonstrate a path forward for gathering this information using large-scale phone surveys that rely on random sampling and standardized questions for comparability across settings. The methodology and harmonized measurement tools can readily be rolled out in new contexts to cover additional populations during this and future crises.
Following on decades of steadily increasing incomes across major world regions, the sharp rise in global poverty in 2020 that we document is unprecedented. The median proportion of respondents across our sample countries experiencing reduced income is a staggering 67%, and negative effects are experienced by households across the socioeconomic spectrum. The economic distress caused by the COVID-19 pandemic has had an immediate cost in terms of nutrition in LMICs. In addition to direct health consequences, hunger places long-run productivity and growth at risk as households compensate by reducing other investments in productive inputs such as fertilizer, selling productive assets, and lowering investment into long-run child development and education. Evidence abounds that these severe shocks to food security of children can threaten long-term health and well-being.
Humanitarian relief efforts that aim to address these problems face two added complications during the pandemic, relative to standard relief programs during regular years. The first is that further viral spread is fundamentally linked to the extent of economic deprivation, and successful disease containment requires the provision of immediate economic relief. Second, the worldwide financing of largescale relief is constrained by the aggregate nature of the COVID crisis that simultaneously affected donor countries, as well as the large magnitude of the global economic recession.
The findings in our data highlight the first challenge. Households facing acute food shortages may be less willing to adhere to social distancing rules than others and could instead seek out income-generating opportunities even in crowded and epidemiologically risky markets. For social distancing to succeed, people must feel sufficiently secure from deprivation and hunger.
Relief programs should be carefully designed to avoid unintended adverse public health consequences-such as increased face-to-face market transactions in areas with high likelihood of viral spread. Cash or food transfers that allay this direct need could even double as tools to address disease spread by discouraging such market interactions. For example, transfers could be explicitly labeled with a "soft" form of conditionality, such as "this is money for food to reduce your need to work in crowded markets," to further promote social distancing. Furthermore, new innovations to quickly and safely identify the poor using mobile phones or satellite data [e.g.,] and deliver funds remotely through mobile money transfershold promise in this context because of the minimal contact required to implement.
Our data also highlight the widespread nature of the global economic shock. Social protection programs in LMICs are underfunded even in good times. During an economic downturn, reduced tax revenue will make financing such programs even harder, and debt markets are not readily available for LMICs. Because the severity of the current crisis makes it important to expand safety net programs, international support-for instance, in the form of grants or concessionary Monthly rates of food insecurity among landless agricultural households in northern Bangladesh from sample BGD5. Food insecurity is defined as missing a meal or reducing portions for at least 15 days in a month. Note that this is a more stringent criterion than that reported in; in this figure, we restrict to cases of frequently missed meals. The 2020 rates come from an April phone survey, and "Previous year" reflects retrospective survey data spanning January 2018 through May 2019 collected in two survey rounds in February and June 2019. (B) Data from agricultural households in western Terai, Nepal, from sample NPL1. The index of food insecurity is constructed using two questions on how often households had to worry about not having enough food or had to reduce portion sizes. The data points in late 2019 and early 2020 come from six rounds of contemporaneous phone survey, and "Previous year" reflects respondents' recollection about a prior "typical year" reported during the April-May 2020 phone survey round. loans-will be needed. Rich countries that are themselves under pressure from this same health and economic crisis may be tempted to focus on addressing problems at home. Yet, since disease transmission does not respect national borders, it is in the self-interest of wealthy countries to help reduce the spread of COVID-19 in LMICs, over and above any humanitarian motivations.
Policymakers in LMICs will also need to craft creative solutions to develop income-generating activities with longer gestation periods in case the risky COVID-19 disease environment or the associated economic slowdown persists for a prolonged period. For instance, "graduation programs" that combine assets and training can promote a source of livelihood that requires limited external contact and have been shown to reduce poverty in the past. Combining these programs with immediate cash support has even been shown to help build sustainable sources of income during periods of civil unrest [e.g.,].
On an optimistic note, the innovation and technological adoption that takes place during emergencies can spur long-run economic development. Dealing with the economic fallout from COVID-19 will require the technological infrastructure to reach poor populations in remote areas with minimal face-to-face contact. These workarounds have accelerated the expansion of new financial technology during past political and economic crises [e.g.,]. Solutions that arise in the current climate thus have the potential to both improve resilience immediately and durably advance the financial ecosystem.
Countries around the world face difficult policy choices along the path to economic recovery from COVID-19. While much public discussion focuses on "lives" and "livelihoods," our data suggest this is a false dichotomy. We provide systematic evidence on how the outbreak has adversely affected households across multiple LMICs in several major world regions. A more appropriate framing of the situation in these countries could be in terms of "lives damaged or lost due to disease" and "lives damaged or lost due to economic deprivation." We emphasize that our data do not speak to the economic consequences of imposing or relaxing specific lockdown policies. However, the evidence does have specific policy implications for how to cope with the economic hardships, to protect both lives now and in the future: fund and implement immediate humanitarian relief and long-term safety net programs to ameliorate the damage that we document.
# Supplementary materials
|
Multiplex immunofluorescence is a powerful tool for the simultaneous detection of tissue-based biomarkers, revolutionising traditional immunohistochemistry. The Opal methodology allows up to eight biomarkers to be measured concomitantly without cross-reactivity, permitting identification of different cell populations within the tumour microenvironment. In this study, we aimed to validate a multiplex immunofluorescence workflow in two complementary multiplex panels and evaluate the tumour immune microenvironment in colorectal cancer (CRC) formalin-fixed paraffin-embedded tissue. We stained CRC and tonsil samples using Opal multiplex immunofluorescence on a Leica BOND RX immunostainer. We then acquired images on an Akoya Vectra Polaris and performed multispectral unmixing using INFORM. Antibody panels were validated on tissue microarray sections containing cores from six normal tissue types, using QUPATH for image analysis. Comparisons between chromogenic immunohistochemistry and multiplex immunofluorescence on consecutive sections from the same tissue microarray showed significant correlation (r s > 0.9, Pvalue < 0.0001), validating both panels. We identified many factors that influenced the quality of the acquired fluorescent images, including biomarker co-expression, staining order, Opal-antibody pairing, sample thickness, multispectral unmixing and biomarker detection order during image analysis. Overall, we report the optimisation and validation of a multiplex immunofluorescence process, from staining to image analysis, ensuring assay robustness. Our multiplex immunofluorescence protocols permit the accurate detection of multiple immune markers in various tissue types, using a workflow that enables rapid processing of samples, above and beyond previous workflows.
Multiplex immunofluorescence is a powerful tool for the simultaneous detection of tissue-based biomarkers, revolutionising traditional immunohistochemistry. The Opal methodology allows up to eight biomarkers to be measured concomitantly without cross-reactivity, permitting identification of different cell populations within the tumour microenvironment. In this study, we aimed to validate a multiplex immunofluorescence workflow in two complementary multiplex panels and evaluate the tumour immune microenvironment in colorectal cancer (CRC) formalin-fixed paraffin-embedded tissue. We stained CRC and tonsil samples using Opal multiplex immunofluorescence on a Leica BOND RX immunostainer. We then acquired images on an Akoya Vectra Polaris and performed multispectral unmixing using INFORM. Antibody panels were validated on tissue microarray sections containing cores from six normal tissue types, using QUPATH for image analysis. Comparisons between chromogenic immunohistochemistry and multiplex immunofluorescence on consecutive sections from the same tissue microarray showed significant correlation (r s > 0.9, Pvalue < 0.0001), validating both panels. We identified many factors that influenced the quality of the acquired fluorescent images, including biomarker co-expression, staining order, Opal-antibody pairing, sample thickness, multispectral unmixing and biomarker detection order during image analysis. Overall, we report the optimisation and validation of a multiplex immunofluorescence process, from staining to image analysis, ensuring assay robustness. Our multiplex immunofluorescence protocols permit the accurate detection of multiple immune markers in various tissue types, using a workflow that enables rapid processing of samples, above and beyond previous workflows.
# Introduction
Opal multiplex immunofluorescence (mIF) is based on the principle of tyramide signal amplification (TSA), wherein fluorescent Opal dyes are conjugated with tyramide molecules to produce enzymatic amplification, similar to conventional chromogenic immunohistochemistry (IHC). This enables investigators to more accurately detect the presence of infrequently expressed biomarkers and permits the use of well-validated primary antibodies raised in the same species. Application of these assays was previously limited to small-scale studies since quantitative analysis of wholeslide mIF staining in large cohorts was rendered unfeasible by the lack of assay automation, whole-slide imaging and dedicated image analysis software. Today, however, there exist autostainers capable of automating mIF protocols with quick turnaround times. Such protocols would have taken up to a week to conduct manually, increasing the likelihood of batch effects and human error. In addition, slide scanners now offer the possibility of fluorescent wholeslide image capture. The resultant multilayered pyramidal tiff files can be easily processed using commercial and open-source software that are now available for comprehensive digital image analysis (DIA).
The ability to use standardised DIA algorithms for the assessment of chromogenic IHC staining has been markedly improved by the use of validated, automated staining protocols, standardised whole-slide scanning and image processing. In essence, optimisation of the entire pathology workflow is key in determining assay robustness and troubleshooting efficiency. The possibility of validating automated mIF assays therefore brings the prospect of deploying the technique into clinical practice. However, it is not yet known how considerations during multiplex optimisation, such as Opal-antibody pairings and multispectral unmixing, affect the standardisation of image analysis and digital pathology.
In this study, we describe our experience validating a digital pathology workflow for which two complementary mIF assays were developed. We evaluate how mIF compares to traditional IHC for biomarker quantification, following assay optimisation based on relative epitope stability, optimal antibody concentrations, antibody staining order, Opal-antibody pairings and denaturing protocols. Furthermore, we assess whether factors affecting mIF image acquisition, such as co-expressing biomarkers, crosstalk and relative fluorophore intensity, need to be considered when standardising DIA algorithms. The two multiplex panels were therefore designed to include co-expressing biomarkers to assess the impact of cellular co-expression on DIA. Multiplex panel one (MP1) was designed as a 6-plex assay, comprising CD3, CD4, CD8, CD20, cytokeratin (CK) and DAPI, with CD3 expected to colocalise with CD4 and CD8 in the membrane of T cells. Multiplex panel 2 (MP2) was created as a 5-plex assay, consisting of CD4, CD68, FOXP3, CK and DAPI. This time assesses the effect of two biomarkers, CD4 and FOXP3, colocalising in different cellular compartments of regulatory T cells. In addition, we investigate the importance of pre-analytical factors in DIA such as sample thickness, multispectral unmixing and batch versus individual scan exposures to assess how they may affect the acquired images and subsequent analysis. Using these data, we conclusively determine how important the technical challenges unique to mIF can influence the robustness of a DIA protocol.
# Materials and methods
## Patient samples
The study was conducted according to the Good Clinical Practice guidelines and the Declaration of Helsinki. All patients provided informed consent for sampling of their tissue. Protocol optimisation was carried out on full-face tonsil and colorectal resection specimens provided as 4 µm, 10% neutrally buffered, formalinfixed paraffin-embedded (FFPE) tissue sections. Access to these slides was granted under the NIB approval (OREC 16/NI/0030; NIB15/0168). Ethical approval for use of these samples in research was granted through the Northern Ireland Biobank (NIB15/0168). The optimised mIF protocols were validated against chromogenic singleplex protocols on consecutive sections of a tissue microarray (TMA). The TMA was constructed with 4 9 0.6 mm tissue cores taken from normal tonsil, placenta, colonic epithelium, lymph node, appendix and spleen using a Beecher manual arrayer (Beecher Instruments Inc., Sun Prairie, WI, USA). Both validated mIF protocols were then tested on full-face resection specimens from the tissue of interest, colorectal cancer (CRC). Access to these slides was granted under collaboration with the Stratification for Colorectal Cancer Consortium (S:CORT) with appropriate approvals in place (OREC 15/EE/0241). Ethical approval for use of these samples in research was granted through NHS REC proportionate review (OREC 15/EE/0241). All work on the tissue sections was performed in the Precision Medicine Centre of Excellence at Queen's University Belfast using standardised operating procedures for IHC staining, digital slide scanning and DIA. All procedures were reviewed and agreed by senior consultant pathologists (JAJ and MS-T).
## Tissue preparation for staining
Prior to staining, all tissue slides were deparaffinised on the Leica BOND RX automated immunostainer (Leica Microsystems, Milton Keynes, UK) by baking for 30 min at 60°C, soaking in BOND Dewax Solution at 72°C and then rehydrating in ethanol.
## Chromogenic singleplex immunohistochemistry
Singleplex IHC using 3,3 0 -diaminobenzidine (DAB) detection (BOND Polymer Refine Detection, Leica Biosystems, Newcastle Upon Tyne, UK; Catalogue No. DS9800) was carried out to determine the conditions and the order in which the primary antibodies would be applied in the multiplex protocol. To ensure adequate epitope stability following successive rounds of heat-induced epitope retrieval (HIER), chromogenic singleplex IHC was conducted in the first, intermediate and last round of HIER for each of the biomarkers to be multiplexed in MP1, corresponding to positions 1, 3 and 5 of antibody staining (Data S1). Results of positions 3 and 5 reflected those of positions 2 and 4, respectively, which permitted rapid IHC optimisation. Staining was performed on the Leica BOND RX with HIER pretreatments applied at 95°C using BOND Epitope Retrieval (ER) Solutions: citrate-based pH 6.0 ER1 (Leica Biosystems; Catalogue No. AR9961) or EDTA-based pH 9.0 ER2 (Leica Biosystems; Catalogue No. AR9640).
## Singleplex and multiplex immunofluorescence
The TSA-based Opal method was used in this study for immunofluorescence (IF) staining (Opal Polaris 7-Color Automation IHC Kit; Akoya Biosciences, Marlborough, MA, USA; Catalogue No. NEL871001KT). Since TSA and DAB oxidation are both peroxidasemediated reactions, the primary antibody conditions and order of staining determined using DAB detection were directly applied to the fluorescent assays, using Opal reagents in lieu of DAB IHC reagents (Data S2). Unlike conventional IHC wherein a chromogenic peroxidase substrate is used for antigen detection, each antibody is paired with an individual Opal fluorophore for visualisation. Optimal Opal-antibody pairings were assigned based on expected co-expression and abundance of the biomarkers in CRC tissue. More explicitly, if biomarkers were expected to colocalise in the same cellular compartment, then they were paired with spectrally separated Opals. Additionally, low-expressing markers were coupled to more intense Opals to facilitate spectral acquisition, and vice versa. The Opal fluorophores were used at a 1 in 150 dilution, as recommended by Akoya when using the Leica BOND RX. As such, a fluorescent singleplex was performed for each biomarker and compared to the appropriate chromogenic singleplex to assess staining performance. To confirm the absence of signal crosstalk, a dropout control was ran for each antibody using the final optimised multiplex.
During singleplex development and multiplex optimisation, Opal-antibody pairings, concentrations and denaturing parameters for each biomarker were assessed and adjusted for. This was done by checking the signal-to-background ratio (signal intensity of positive staining: background > 10 : 1) and signal balance (signal intensity of all fluorophores < 30 counts) with Akoya's INFORM software version 2.4.6. We aimed at obtaining the ideal signal intensity range of 20-25 counts for each antibody, which translates as 100-125 nm of fluorescence capture on the Vectra Polaris. All Opal dyes were initially considered for Opalantibody pairings, but along the optimisation process we transitioned to using only MOTiF Opals to take advantage of the MOTiF technology that enables rapid 7-colour whole-slide multispectral imaging of fluorescent slides.
## Image acquisition
All chromogenic immunostained slides were scanned using an Aperio AT2 (Leica Biosystems, Vista, CA, USA) at 409 magnification and then were independently reviewed for quality and consistency by a trained senior technician (VB) before being considered for DIA.
All fluorescently labelled slides were scanned on the Vectra Polaris at 209 magnification using appropriate exposure times. Initially, whole-slide images were scanned with all five standard epi-fluorescence filters (DAPI, FITC, Cy3, Texas Red and Cy5). Then, when MOTiF Opals were solely used, images were acquired using tile scanning with the 7-colour whole-slide unmixing filters (DAPI + Opal 570/690, Opal 480/620/ 780 and Opal 520). Library slides were generated from representative tissue sections to allow for accurate unmixing of the multiplexed samples, including a slide stained for each single fluorophore, a DAPI only slide and an autofluorescence slide wherein no antibody, Opal reagent or DAPI was applied. The unmixing performance of this tissue-specific spectral library was compared to that of the synthetic Opal library available in INFORM. Resultant image tiles were then stitched together within QUPATH v0.2.0-m4 using a script available on the QUPATH GitHub to produce a whole-slide multichannel, pyramidal OME-TIFF image for DIA.
# Digital image analysis
Assessment of all biomarkers was undertaken using the open-source software QUPATH v0.2.0-m4. Each core of the TMA, used for validation, was given a unique identifier in order to directly compare cellular expression between slides. The full-face sections, used for optimisation and testing, were annotated with the assistance of a senior consultant pathologist (MS-T). Following tissue annotation, cell segmentation was carried out based on haematoxylin if chromogenic or DAPI if fluorescent. Phenotyping of the cells was then performed using binary classification in the chromogenic images, or using a bespoke script enabling multimarker cataloguing of detected cells in the fluorescent images. All analysed images were independently reviewed for quality control purposes, with at least 20% being reviewed by a pathologist prior to data export.
During the progress of this study, the latest QUPATH milestone v0.2.0-m9 was released, offering a new approach for multiplex analysis. We analysed the multiplexed TMA sections, used for validation, in this latest version of QUPATH, which provides two methods of mIF classification. Method 1 involves simple thresholding of a single measurement to classify each biomarker, similar to the approach used in v0.2.0-m4. Method 2 utilises a machine learning classifier, random forest, which is trained on multiple measurements. Once the classifiers have been established for each biomarker using either method 1 or 2, these are combined and applied sequentially. The newly generated phenotype results were compared to those previously obtained with QUPATH v0.2.0-m4.
# Statistical analysis
All statistical analyses were conducted using PRISM version 5 (Graphpad Software, San Diego, CA, USA). The relationship between the staining parameters (antibody order, Opal-antibody pairing) and the percentage of positive cells detected was assessed using the Mann-Whitney U-test. For validation purposes, a difference in cell positivity of greater than 10% (i.e., less than 90% accuracy of obtaining a similar result) between DAB and IF cell counts was considered significant. The monotonic relationship between the singleplex IHC and mIF was assessed using Spearman's rank correlation. P-values of less than 0.05 were considered significant. The agreement between the singleplex IHC and mIF assays was assessed using a Bland-Altman plot, and the determinant bias was considered acceptable if within the 95% limits of agreement. Comparison of the QUPATH v0.2.0-m4 and -m9 data was conducted using a paired t-test, with a two-tailed Pvalue of less than 0.05 defined as being significant.
# Results
## Chromogenic singleplex optimisation
Singleplex development is fundamental in designing a multiplex assay as it determines the staining parameters for each individual biomarker in the multiplex panel. In order to identify the best placement of each biomarker in the staining sequence for MP1, three chromogenic IHC singleplexes for each primary antibody were performed on serial sections of low and high immune expressing CRC tissue. Epitope stability was evaluated as a function of the number of HIER cycles preceding antibody application. Stability was defined by the absence of relative change in cell positivity across the serial sections. Epitope denaturation, on the other hand, was represented by a decline in staining intensity and consequently a drop in positive cells detected. In the current study, CD20 was found to be an epitope requiring minimal retrieval and was therefore best placed in position 1. Results also showed CD3 required limited retrieval and could be placed in position 2, whereas CD4, CD8 and CK exhibited an increase in cell positivity with continued epitope retrieval, and were therefore best placed in positions 3-5.
## Fluorescent singleplex optimisation
Based on the visual assessment and quantitative analysis of the chromogenic singleplex stains, the preliminary order of MP1 staining consisted of CD20 in position 1, followed by CD3, CK, CD8 and CD4. Taking into consideration biomarker co-expression and relative abundance, initial Opal-antibody pairings were assigned to each biomarker based on the chromogenic singleplex results. CD4 and CD8 were coupled to fluorophores spectrally distant from the CD3-associated Opal, in addition to them being sequentially separated from CD3 (permitted by assigning CK to position 3). CD3, CD4, CD8 and CK were all highly expressing targets in the representative tissue type, so they were coupled to low-or medium-intensity Opals. CD20, which was found to be low expressing, was paired with a brighter fluorophore (Opal 650).
The fluorescent singleplexes were performed on the same low and high immune expressing CRC cases as the chromogenic singleplexes to permit comparison. Positive cell detection in similar regions of interest (ROI) demonstrated less than 10% difference in all but four cases: CD3, CD8 and CK in high immune CRC; CK in low immune CRC. These significant differences are reflective of the variant cellular profiles that exist among nonserial sections. Review of these cases found that such differences were also attributed to QUPATH's inability to accurately quantify cells in regions with dense populations of DAB-positive cells. DAB is an opaque stain; therefore, colour deconvolution in these intensely stained samples was inept in separating the haematoxylin channel for accurate cell detection.
## Multiplex immunofluorescence optimisation
With optimal staining parameters established for each biomarker, creation of a spectral unmixing library, imperative to the accurate analysis of the forthcoming multiplex slides, was required. While the manufacturer's guide recommends building library slides specific to the tissue type of the study, we anticipated many amendments to the Opal-antibody pairings during multiplex optimisation, which would require a new library to be built each time. As a means to replace this step, we compared the spectra and unmixing performances of our newly created CRC-specific library and the synthetic Opal library (constructed from tonsil tissue) (Data S3). No significant differences were observed, thereby validating the use of the synthetic library moving forward.
The fluorescent singleplex protocols were combined into an initial multiplex protocol, which was progressively modified following the manifestation of spectral bleed through. Most predominantly, CD20 crosstalk was observed in the CD8 channel, in both CRC and tonsil tissue. This was initially explained by the use of CD20-Opal 650 and CD8-Opal 690 pairings as both Opals are acquired in the Cy5 channel, within which their emission peaks overlap (Data S3). By changing the Opal-antibody pairings, the possibility of spectral crosstalk was eliminated, yet crossover remained present in immune hot spots, that is germinal centres. Dropout controls confirmed that the observed crosstalk was not due to inadequate antibody stripping (Data S4). As CD8 was placed in position 3 of MP1, we determined that epitope instability may be inducing this effect. However, due to the inclusion of less stable HIER antibodies CD20 and CD3 in MP1, we could not modify the staining position of CD8. The other issue originally observed was CK bleed through into CD8 and CD4 channels, which was easily addressed by reordering the biomarkers.
In addition to revising Opal-antibody pairings, positions and dilutions, there were two major changes made during mIF protocol optimisation. The first was a switch to using solely ER1 for antibody stripping. Because ER1 is a gentler epitope retrieval method, using a pH of 6 as opposed to a pH of 9 in ER2, it generates less nonspecific staining and background signal. The second amendment was a switch to using only MOTiF Opals for signal amplification. MOTiF Opals are a preferred choice as they allow scanning times 209 quicker than was formerly possible, owing to the fact that they are excited by filters located on merely three filter wheels in the Vectra Polaris (only available with the upgraded filter cube for 7-colour whole-slide unmixing). Moreover, MOTiF Opals have spectrally distant peaks that permit efficient multispectral unmixing.
Subsequent to each adjustment made in the protocol, signal-to-background ratio and signal balance for each revisited Opal-antibody pairing were reassessed. Optimal signal intensities of 20-25 counts were achieved as the protocol developed.shows example images of a CRC TMA core stained with one of these later protocols, andpresents the final. Columns 6-8 present the spectral properties of the fluorophores and the last column the fluorophores that qualify for MOTiF scanning. (B) Fluorescent singleplexes (n = 10) were compared to the chromogenic singleplexes (n = 10) previously performed on the same low and high immune expressing CRC tissue. Bar graph demonstrating the percentage of cell positivity for each biomarker in low and high immune expressing CRC tissue when stained with DAB (brown) versus Opal (multicolour) reagents. Statistical significance was determined by the Mann-Whitney U-test. Difference of > 10% in cell positivity in four cases, marked with an asterisk. (C) ROI of two of these cases is displayed. For each case, the chromogenic stain is seen on the left and fluorescent stain on the right, with the original image above and with cell detection applied below. The CD3 stains are viewed at 209 magnification (scale bar = 50 µm), and the CK stains are seen at 109 magnification (scale bar = 100 µm). In the chromogenic CD3 stain, there are more positive cells (in brown) than there are in the fluorescent CD3 stain (in yellow), which is reflective of the different immune expression profiles that exist between nonserial sections. In the chromogenic CK stain, some DAB-positive cells are not detected by QUPATH as being a negative cell (blue) nor a positive cell (red). However, all cells are identified in the fluorescent CK stain as being either negative (grey) or positive (green).
## Optimisation of high-throughput image acquisition
The Vectra Polaris allows high-speed digital wholeslide scanning at 109, 209 and 409 magnification in both brightfield and fluorescence. Once the final multiplex protocols (MP1 and MP2) were realised, 209 and 409 fluorescent scanning were tested on a MP1stained CRC full-face sample. The high resolution in the 409 scan offered a more defined image of the cells and their surface boundaries than the 209 scan. However, there was less than 1% difference in the cell populations detected and such detail was not considered essential for the purpose of this validation study . So with the advantage of minimised scanning time, 209 scanning was selected hereafter. Spearman's rank correlation determined that no difference exists in their staining intensities. (C) Two serial full-face CRC sections were stained, one for MP1 (top row) and one for MP2 (bottom row). The difference in their staining intensities is seen in the left column: CK (green) and CD4 (cyan), which are present in both panels, are more intense in MP2 than in MP1. This originates from the difference in their paraffin section thickness, seen in the right column via the autofluorescence channel: Autofluorescence in MP1 is greater than in MP2 on account of the MP1 section being thicker than the MP2 section. Images are viewed at 59 magnification (scale bar = 250 µm). (D) A colonic core before (top row) and after (bottom row) spectral unmixing in INFORM. The epithelium is expected to stain green for CK (Opal 480) and not yellow for CD3 (Opal 520). Other MP1 markers seen here are CD20 in red (Opal 570), CD4 in cyan (Opal 620) and CD8 in magenta (Opal 690). Images are viewed at 109 magnification (scale bar = 100 µm).
When setting exposure times for image acquisition, the manufacturer's guide recommends autoexposing on several slides in a batch for each fluorophore. This method of batch exposure was compared to individual exposure, wherein the channels were autoexposed for each individual slide, thus generating slide-specific scanning protocols. Once scanned, a single autofluorescence image was used for spectral unmixing. Results revealed no difference in staining intensities, both visually and quantitatively (r s > 0.9, P < 0.05). So with it requiring less preparation time, batch exposure was applied moving forward.
Sample thickness, on the other hand, was found to be a determining factor in producing reliable staining intensities. Although tissue blocks were cut with the intention of generating regular 4-µm sections, we encountered intersection thickness variability during mIF optimisation, which occurs as a result of thermal expansion in microtomy. For instance, two serial full-face CRC sections were stained with MP1 and MP2. Both comprising of CD4 and CK, the panels were expected to produce similar expression patterns for helper T cells and epithelial tumour cells. However, staining was considerably weaker in the MP1 scan compared to the MP2 scan, which was due to the MP1-stained tissue being thicker than the MP2stained tissue. Since FFPE tissue is inherent to autofluorescence, and autofluorescence signal intensity is positively correlated with tissue thickness, viewing the images in the autofluorescence channel enabled visual comparison of their section thickness. The MP1-stained section was notably brighter in autofluorescence intensity and hence thicker. Evidently, regular thin cuts of tissue with a thickness of 3-4 µm are considered most suitable for immunofluorescence.
Following image acquisition, it was essential to import the scanned images into INFORM for spectral unmixing, which included removing autofluorescence using a representative autofluorescence spectrum. We tested the importance of unmixing by importing scans directly into QUPATH from the scanner. The channels remained mixed and the autofluorescence present. As a result, there were high levels of nonspecific staining, which rendered the determination of certain biomarker thresholds and subsequent cell classification impossible. As seen in, the most apparent nonspecific staining was associated with Opal 520 (assigned to CD3), which emits in the FITC channel. Consequently, the colonic epithelium appears to be staining for CD3. As an alternative workflow, the same scans were first opened in the whole-slide viewer PHENOCHART v1.0.12 (Akoya Biosciences, Marlborough, MA, USA) to select for ROI, imported into INFORM for unmixing using the synthetic library and then stitched together in QUPATH. The same example image, now unmixed, no longer appears to be staining for CD3 in the epithelium. The substantial difference in image quality therefore confirms that this workflow is necessary for DIA in QUPATH.
## Validation of high-throughput multiplex immunofluorescence protocols
Having fully optimised two panels and implemented a streamline workflow from staining to acquiring quality digital images, the next stage was to validate the protocols and establish a process for analysing the fluorescent images. In the first instance, a TMA block comprising of six normal tissue types was cut into nine consecutive sections. Of these, seven sections were chromogenically stained with a single antibody, one slide for each different biomarker, leaving two sections to be stained with the optimised multiplex panels. The chromogenic TMAs were annotated in QUPATH, and cell detection was carried out using appropriate DAB thresholds. The multiplexed TMAs were annotated in the same way, mimicking the chromogenic ROI as much as possible. Cell classification was done by meticulously determining thresholds for each marker, analogous to the DAB thresholds. Each phenotype was initially identified as a 'singleplex detection', wherein one marker was exclusively detected across the entire multiplexed TMA, for example CD20-positive cell detection on MP1 TMA. These 'singleplex' results were compared to the DAB results of the consecutive sections to assess biomarker correlation and biasnessfor MP1; Data S7 for MP2).
Cell detection results of all the chromogenic TMAs (sections 1, 2, 6 and 7) correlated significantly with the singleplex detection data from MP1 TMA (section 4) (r s > 0.9, P < 0.0001), except for CD8 (TMA section 3). With the latter being nearest to the MP1 TMA section in terms of depth level in the original TMA block, one would expect their immune expression profiles and therefore their detection results to be most comparable. Nonetheless, the poor correlation and high biasness appeared to be driven by three tissue types: spleen, tonsil and lymph node. In spleen, the histiocytic littoral cells that are highly positive for CD8 were detected by DAB staining but not by mIF. In tonsil and lymph node tissue, the CD8 positivity associated with CD20 + germinal centres was picked up by mIF but not by conventional IHC, supporting our conclusion of reduced epitope stability . Similarly, cell detection results for MP2 (TMA section 5) displayed strong correlation for all but one biomarker, FOXP3 (TMA section 8) (P > 0.05), which is likely due to the protein being scarcely expressed in the TMA (< 0.5% expression). Meanwhile, any biasness observed among the other biomarkers was explained by the occurrence of intense DAB staining affecting accurate cell detection. Overall, all determinant biases fell within the limits of agreement and there was significant positive correlation between the DAB and mIF stains, thus validating the multiplex protocols.
## Optimisation of digital image analysis protocols
As well as running 'singleplex detection' on the multiplexed TMAs, cells were classified for all the phenotypes in a hierarchical manner. During this process, we found that the order of classification considerably influenced the phenotyping results for MP1, but not for MP2. Six detection orders were tested on MP1 TMA 4and four were tried on MP2 TMA 5 (Data S9). As seen in, detecting strongly expressing targets early on in the classification order masks the subsequent detection of weaker, neighbouring biomarkers. This was particularly evident in immune hot spots such as germinal centres. By detecting membranous CD20 staining first, recognition of other membranous lymphocyte markers was hindered. As a result, CD20 was most suitable towards the end of the order but still placed before CD8 to avoid true B lymphocytes being classed as CD8 + cells, considering the nonspecific CD20-CD8 staining present in these lymphoid follicles. In order to facilitate the multimarker cataloguing of cells in these multiplexed TMAs, a script was created for each panel (Data S10 and S11). These scripts translate as decision trees; Data S12). For each panel, unexpected classes of cells were identified and integrated into the script. The original script identifying only expected phenotypes was compared to the updated script incorporating unexpected phenotypes, and differences in cell classification were only observed for MP1. Such differences were seen exclusively in the TMA cores containing lymphoid tissue, wherein a portion (up to 16%) of CD4 + cells was reclassified as CD4 + /CD8 + cells. All of these cells were located within groups of immune cells; none were isolated CD4 + cells.
Multimarker classification was also performed using QUPATH v0.2.0-m9 to compare its novel multiplexed analysis approach (methods 1 and 2) to our current hierarchical method. Results of this analysis determined that cell classification in v0.2.0-m9 intuitively enables multiclass cell labelling, rendering the cell detection order extraneous. Comparison of multiclass labelled cells (in v0.2.0-m9) to cells classified by our hierarchical method (in v0.2.0-m4) for MP1 found that, in the absence of hierarchical structuring, cells in tissues containing tertiary lymphocytic structures were likely to be further subcategorised based on relative proximity to CD20 + cells (Data S13). Using v0.2.0-m9 method 1, significant differences in cell positivity were observed for CD3 À /CD4 + and CD3 + /CD8 + phenotypes in lymph node tissue (P-values of 0.0173 and 0.0162, respectively). Method 2 introduced even greater differences across multiple cell phenotypes (CD20 + , CD3 À / CD8 + , CD3 + /CD8 + , CD3 + and CK + ; P-values < 0.05) and tissue types, deriving from the over-representation of CD20 + cells as well as the use of machine learning. Similar differences were observed for MP2; only this time, CD4 + cells were being further categorised by CD68 positivity. However, the percentage of CD4 + / CD68 + cells was only statistically significant in tissue with high CD68 expression, spleen with method 1 and lymph node with method 2 (P-values of 0.0212 and 0.0275, respectively). Since CD4 constitutes a biomarker in both panels, whose detection differed between v0.2.0-m4 and v0.2.0-m9, we compared the three multiplexed analyses of CD4 to the CD4 DAB data across all tissue types. Classification accuracy decreased from v0.
## Testing the validated workflow: from staining to digital image analysis
For completion of the study, the validated mIF protocols were performed on 10 full-face CRC patient samples (five for MP1 and five for MP2). Validity of successful multispectral unmixing was assessed using the autofluorescence channel. Consequently, three sections were excluded from analysis due to levels of autofluorescence exceeding the spectral range used for autofluorescence subtraction, causing incomplete spectral unmixing. The v0.2.0-m4 scripts were applied for classification of both expected and unexpected phenotypes. MP2-stained sections exhibited no differences between scripts with or without unexpected classes, similarly to the TMA. Likewise with the TMA, MP1stained samples demonstrated a difference in phenotype results between the two scripts, with up to 5% of CD4 + cells being recategorised as CD4 + /CD8 + cells within immune dense regions, thus highlighting the need for binary hierarchical classification.
# Discussion
This study conclusively demonstrates the reliability and credibility of mIF in assessing immune biomarkers using the quantitative DIA QUPATH. mIF is unique in its ability to detect both the expression and geographic cellular distribution of immune biomarkers while preserving tissue architecture, contrarily to other techniques such as flow cytometry. Herein, we present the first study to clearly define and demonstrate the technical limitations that exist for DIA of mIF tissue samples, more specifically, mIF using Opal fluorophores to detect co-expressing immune biomarkers in the same cellular compartment.
For both of our mIF panels, all immunophenotypes were confirmed using singleplex DAB IHC and any steric hindrance or spectral crosstalk was ruled out using dropout controls. MP1 included three biomarkers (CD3, CD4 and CD8) that colocalised in the same cellular compartment of the same cell, whereas MP2 consisted of two biomarkers that co-expressed in different cellular compartments (CD4 and FOXP3). Both panels included CD4 and CK, allowing panel design comparison. Using normal human tissue types, we identified the importance of the detection order in image analysis when assessing sites of densely populated co-expressing immune cells, such as lymph node and tonsil. In these regions, cell detection was predisposed to overestimating whichever immune cell type was initially classified in the sequential binary classifier. This is because biomarkers that share very close subcellular locations can hinder antibody binding, signal amplification and digital detection of their co-expressers by steric hindrance and masking effects. As a result, incompatible immunophenotypes were obtained when simultaneously classifying multiple biomarkers in tissue stained with MP1, but not with MP2. Incompatibility may arise as a result of overlapping cells in thick tissue sections or as a consequence of minimal focal planes in thin sections. However, these factors were ruled out by the acceptable autofluorescent levels present in the validation TMA sections and the concordance observed for each biomarker between the paired singleplex DAB IHC and the mIF stains when binary phenotyping was used, in agreement with pathological review. In utilising this method, comparable CD4 and CK results were obtained in MP1 and MP2. These findings are in accordance with recent publications in this area, wherein binary phenotyping has been reported as being superior to multiple phenotyping in terms of sensitivity in detecting true unexpected dual expressions. Interestingly, inclusion of CD3 in the first panel permitted identification of a subpopulation of CD3-positive, CD4/CD8 null immune cells that would otherwise not have been identified. This population of cells is well reported in other multiplexing technologies such as flow cytometry.
The importance of sequential, binary and overall hierarchical classification is reinforced by our v0.2.0-m9 data. Although the classifiers were combined in a specific order, the cell classification in v0.2.0-m9 was in fact nonhierarchical. To bypass this issue while the QUPATH milestone remains a work in progress, the additional classes can be reclassified into their respective true phenotypes using bespoke scripting, which we did not apply for the sole purpose of differentiating the effects of binary and multiple phenotyping. Moreover, our findings emphasise the need to understand the expected immunophenotypes when carrying out DIA. CD4/CD8 dual positivity, for instance, was accepted as true based on current literature. CD4/ CD68 double positivity, on the other hand, would require further investigation since this phenotype exists in the literature as a low CD68 expresser. This would encompass reanalysing the DAB and MP2 data with appropriate thresholds for low-and high-expressing CD68 cells. Furthermore, the controversial classes of cells additionally identified in v0.2.0-m9 were more substantially present in method 2 using the random forest machine learning, a method that is highly accurate using limited training data but is liable to systematic errors when applied across large cohorts, and is thus apt to introducing bias data. We therefore deduce thresholding to be a more accurate approach in the current study.
The determination of appropriate and definitive thresholds for each particular biomarker in the multiplex panels was enabled by performing DAB IHC in the initial steps of mIF optimisation. Essentially, each target required wet-lab optimisation for DAB IHC, singleplex IF and the full mIF assay on control tissues to guarantee reproducible digital results. Interestingly, the targets that were most affected by successive epitope retrieval when moving from DAB IHC to mIF have been reported as being sensitive to HIER prior to their use in mIF, a process that requires successive rounds of HIER to remove the previously bound antibody. Among the methods of HIER, we noticed that alkaline-based ER2 (pH 9.0) was prone to inducing high background signal compared to acidic ER1 (pH 6.0), which is supported elsewhere in the literature. Furthermore, by means of DAB IHC we were able to discern the cause of the CD20-CD8 crosstalk present in the germinal centres of the fluorescent images. These staining artefacts arose as a consequence of epitope instability in the multiplex. Decreasing epitope stability in sequential staining has been previously reported and can be mitigated by antibody positioning in the staining sequence. However, in this study we demonstrate that placement in the optimal position may not be possible due to other biomarkers present in the panel, thus resulting in a trade-off in accuracy which can be moderated using selective binary phenotyping.
The multiplex panels in the current study were developed using Opal reagents based on TSA methodology. During optimisation, we showed that the primary incubation times used for DAB IHC were identical for TSA-based IF, thus confirming the use of TSA as a DAB equivalent. TSA is most advantageous for its high sensitivity and specificity in detecting low-expressing biomarkers, and in its ability to minimise the risk of antibody cross-reactivity. In our acquired images, we did not see any nonspecific background staining except for FOXP3 antibody in MP2, where red blood cells exhibited nonspecific FOXP3 reactivity. This phenomenon has been reported elsewhere and is in fact specific to the mIF assay, since no cross-reactivity was observed in the singleplex DAB IHC. Furthermore, it has been shown in the literature that CK staining can be inconsistent between mIF batches compared to DAB IHC. We were able to alleviate this problem and obtain CK mIF staining comparable to the DAB IHC by merely changing the Opal-antibody pairing for CK, coupling it to the brightest Opal dye (Opal 480).
In addition to optimising the staining protocol, the use of multispectral unmixing aided the determination of positive cell detection. By unmixing the individual channels, clean images were produced, free from autofluorescence and crosstalk. Similarly to other studies, we found when utilising FFPE tissue the use of multispectral unmixing was most expedient over traditional IF image capture, because unlike fresh frozen tissue FFPE tissue is inherently autofluorescent. By using the Opal MOTiF reagents in conjunction with the Vectra Polaris, we were able to quickly, reliably and reproducibly generate results from a digital mIF image in under 24 h from time of tissue sectioning. Even without MOTiF technology, the Opal methodology permits high-throughput biomarker analysis. Ideally, mIF would be carried out on fresh frozen tissue sections; however, tissue architecture is not as well preserved in fresh frozen tissue as FFPE. Since most routinely processed and retrospectively collected tissue are FFPE, studies have been conducted to find an optimal long-term fixative suitable for processing tissue for fluorescence. So for mIF to be introduced into the clinical setting, robust automated protocols that can be independently validated are required. Formerly, manual mIF took 3 days at the bench, but now automated protocols can be ran overnight. While assay automation offers speed and reproducibility, errors may still arise, such as the appearance of a staining gradient across a slide. However, this phenomenon was not observed in samples assessed for DIA in the current study, owing to the regular preventative maintenance and calibration of the autostainer as well as the stringent review of all slides considered for DIA during quality control procedures. Thus, with automated whole-slide scanning and DIA, mIF is becoming more amenable in facilitating high-throughput use, as demonstrated in the current study and previous studies. Our further investigation into scanning parameters found that increasing magnification from 209 to 409 did not improve the delineation of individual cellular phenotypes when digitally assessing co-expressing biomarkers in the same cellular compartment. In fact, difficulty in accurately phenotyping densely populated cells in fluorescent histological images has been reported elsewhere, thus supporting our findings and our decision to use quicker scanning times to produce images for high-throughput use in this study. In future, the classical approach of watershed cell detection currently utilised in QUPATH may be superseded by use of deep learning-based methods for nucleus segmentation, particularly when analysing regions of crowded cells. A promising approach would involve incorporating StarDist 2D, a method that localises nuclei via star-convex polygons, into QUPATH by scripting.
An added benefit of using multispectral imaging and collecting in the autofluorescent channel was the determination of relative tissue thickness. We found that the predefined thresholds, based on initial DAB IHC and calibrated to the mIF validation assay, were dependent on very slight changes in tissue thickness. In essence, thresholds were significantly affected when the tissue section being assessed was cut either slightly thicker or thinner than expected. Natural variation in tissue thickness is expected and often seen in routine basic and special stains due to manual sectioning of the tissue. It is a recognised issue in manual IF assessment and one that is becoming increasingly apparent when developing artificial intelligence algorithms in histological images. Compared to brightfield image analysis, assessment of mIF is uniquely sensitive to very subtle changes in tissue thickness, which can either increase or decrease the potential availability of exposed epitopes to fluorophore binding. This can result in underexposure or oversaturation of the fluorophore when imaged with the multispectral scanning protocol, which has been optimised for a particular tissue thickness. Therefore, the use of tissue sections with regular thin cuts is highly recommended for mIF assays.
Colour inconsistency is an issue in histological images that arises from pre-analytical variables such as section thickness, which has become more palpable with the emergence of DIA. Image normalisation has been used with great effect in the digital assessment of brightfield images that display such colour inconsistency. We attempted to apply image normalisation techniques to recalibrate the staining intensity of tissues that exhibited different thicknesses, but found it introduced digital artefacts and rendered the images not fit for purpose (data not shown). Although the IN-FORM software used for multispectral unmixing normalised each image to the reference channels of the synthetic library, it did not account for tissue thickness variation as only one autofluorescent spectrum was provided. A prospective approach to correcting for intersection thickness variability, which was not assessed in this study, would therefore involve using autofluorescent spectra captured from slides of varying thicknesses in combination with the synthetic Opal library. This would enable complete spectral unmixing and more accurate DIA, irrespective of the inherent autofluorescence of FFPE tissue.
# Conclusion
This study not only demonstrates the importance of optimising the wet-lab workflow for the production of quality mIF images, but also highlights the significance of standardising the DIA protocol before undertaking any large mIF studies. Detailed quality control of every step, from staining to analysis, is therefore necessary in ensuring accurate cellular profiling. Herein, we present a digital pathology workflow that allows for automated high-throughput mIF and accurate immunophenotyping. Although there exist novel mIF platforms that offer much larger plex panels than the method used in this study, such as Akoya's CO-Detection by indEXing (CODEX), Nanostring's Digital Spatial Profiling (DSP) and Ultivue's InSituPlex, these incur many limitations that do not permit cost-and time-effective whole-slide imaging. Equally, there are several DIA software that are compatible with most mIF imaging modalities, including Indica Labs' HALO and Visiopharm's Oncotopix. However, unlike QUPATH, these proprietary software do not offer scripting functionalities, a feature considered essential to our study. We also demonstrate the importance of minimising pre-analytical variables for the assessment of the resultant digital image, including biomarker combinations of interest and tissue thickness. Such considerations permitted the creation of a reproducible pipeline for the quantitative assessment of mIF assays in FFPE tissue. Moreover, we found that pathologist supervision during the development of the DIA protocol was invaluable as certain immunophenotyping combinations being interrogated may have otherwise been overlooked. To conclude, this study provides a robust, methodological guide for mIF validation for use in cancer immunology studies.
Halio, consultancy with Nordlai-Adlyte and advisory boards with Merck, Amgen and Roche. Prof Manuel Salto-Tellez is a senior scientific advisor to Philips Computational Pathology and Sonrai Analytics, and has received honoraria from Roche, AstraZeneca, Merck and GSK. These declarations of interest have no relationship with the submitted publication. All other authors declare no competing interests.
# Author contributions
AVP, SGC, VB, KM, MPH, SS, SDR, PQ, LC, ED, TSM, JAJ and MS-T contributed to interpretation of data and writing of the manuscript. AVP, SGC, JAJ and MS-T were involved in study conception and design. AVP, SGC, VB, SDR and LC contributed to data acquisition. AVP, SGC, KM and SS were involved in data analysis.
## Data accessibility
Data are held within the Northern Ireland Biobank and the stratified medicine consortium in colorectal cancer (S:CORT), and are available upon reasonable request.
Data S9. Insignificance of the detection order in digital assessment of MP2. Data S10. Script used for multi-marker classification of cells in MP1. Data S11. Script used for multi-marker classification of cells in MP2.
Data S12. Decision tree visually representing the script used for MP2 cell classification. Data S13. Comparison of MP1 phenotype results obtained in QuPath v0.2.0-m4 and v0.2.0-m9 (method one and two). |
Risk factors for hospitalization and severe outcomes of 2009 pandemic H1N1 influenza in Quebec, Canada
Background/Objective This case-control study was carried out to estimate risk factors associated with hospitalizations and severe outcomes [intensive care unit (ICU) admission or death] among patients with illness because of laboratory-confirmed 2009 pandemic A ⁄ H1N1 virus (pH1N1) during the first wave of pH1N1 activity in the province of Quebec, Canada.Patients/Methods We collected epidemiologic information by phone using a standardized questionnaire from patients with laboratory-confirmed pH1N1 illness during the first spring ⁄ summer pandemic wave in Quebec, Canada. Risk factors associated with hospitalization were assessed by comparing hospitalized to community cases and for ICU admission or death through comparison with hospitalized cases.Results Cases (321 hospitalized patients including 47 ICU admissions and 15 deaths) were compared to controls (395 nonhospitalized patients) by using multivariable logistic regression adjusted for gender, age, education, being a health care worker, smoking, seasonal influenza vaccination, delay to consultation, antiviral use before admission, pregnancy, underlying medical conditions, and obesity. Age <5 years, underlying medical conditions (neuromuscular, cardiac, pulmonary, and renal conditions, diabetes, asthma, and other), and delayed consultation were associated with hospitalization. The strongest association with hospitalization was observed for neuromuscular disorders. Antiviral medication before hospital admission protected against severe disease. Association of obesity with hospitalization was not significant after adjustment in multivariable analysis. Among hospitalized patients, age ‡60 years and immune suppression were associated with death.Conclusions Previously identified risk factors for seasonal influenza were also associated with increased risk of severe pH1N1 outcomes. The independent role of obesity needs to be further defined.
# Introduction
As with seasonal influenza, the vast majority of patients with illness caused by the 2009 pandemic A ⁄ H1N1 virus (pH1N1) experienced an uncomplicated course to recovery. Similar to previous pandemics, the median age of patients presenting with pH1N1 was younger than that of patients with seasonal influenza. [bib_ref] Clinical characteristics and 30-Day outcomes for influenza A 2009 (H1N1), Belongia [/bib_ref] Surveillance reports, outbreak investigations, descriptive case series, and a few case-control studies have attempted to characterize the profile of patients experiencing more severe outcomes. [bib_ref] Clinical characteristics and 30-Day outcomes for influenza A 2009 (H1N1), Belongia [/bib_ref] [bib_ref] Hospitalized Patients with 2009 H1N1 Influenza in the United States, Jain [/bib_ref] [bib_ref] Critically ill patients with 2009 influenza A(H1N1) infection in Canada, Kumar [/bib_ref] [bib_ref] Critically Ill Patients With 2009 Influenza A(H1N1) in Mexico, Dominguez-Cherit [/bib_ref] [bib_ref] Severe respiratory disease concurrent with the circulation of H1N1 influenza, Chowell [/bib_ref] [bib_ref] Pneumonia and respiratory failure from swine-origin influenza A (H1N1) in Mexico, Perez-Padilla [/bib_ref] [bib_ref] Factors associated with death or hospitalization due to pandemic 2009 influenza A(H1N1)..., Louie [/bib_ref] [bib_ref] Infection and death from influenza A H1N1 virus in Mexico: a retrospective..., Echevarria-Zuno [/bib_ref] [bib_ref] Risk factors and outcomes among children admitted to hospital with pandemic H1N1..., O'riordan [/bib_ref] [bib_ref] Mortality from pandemic A ⁄ H1N1 2009 influenza in England: public health..., Donaldson [/bib_ref] [bib_ref] Correlates of severe disease in patients with 2009 pandemic influenza (H1N1) virus..., Zarychanski [/bib_ref] [bib_ref] Risk of severe outcomes among patients admitted to hospital with pandemic (H1N1)..., Campbell [/bib_ref] [bib_ref] Pediatric hospitalizations associated with 2009 pandemic influenza A (H1N1) in Argentina, Libster [/bib_ref] The estimated rate of hospitalization among symptomatic patients with pH1N1 varied by age, ranging from 0AE53% in children 5-17 years of age, 2AE19% in children <5 years of age, [bib_ref] The Severity of Pandemic H1N1 Influenza in the United States, Presanis [/bib_ref] and up to 4AE0% in adults. [bib_ref] Clinical characteristics and 30-Day outcomes for influenza A 2009 (H1N1), Belongia [/bib_ref] Among hospitalized cases, between 6% and 31% were admitted to the intensive care unit (ICU) and up to 5% died. [bib_ref] Hospitalized Patients with 2009 H1N1 Influenza in the United States, Jain [/bib_ref] [bib_ref] Critically ill patients with 2009 influenza A(H1N1) infection in Canada, Kumar [/bib_ref] [bib_ref] Infection and death from influenza A H1N1 virus in Mexico: a retrospective..., Echevarria-Zuno [/bib_ref] [bib_ref] Pediatric hospitalizations associated with 2009 pandemic influenza A (H1N1) in Argentina, Libster [/bib_ref] Secondary bacterial infections have been confirmed in 20-43% of severe cases. [bib_ref] Critically ill patients with 2009 influenza A(H1N1) infection in Canada, Kumar [/bib_ref] Most patients with severe pH1N1 illness had underlying medical conditions known also to increase the risk for complications from seasonal influenza. [bib_ref] Hospitalized Patients with 2009 H1N1 Influenza in the United States, Jain [/bib_ref] [bib_ref] Critically ill patients with 2009 influenza A(H1N1) infection in Canada, Kumar [/bib_ref] [bib_ref] Seasonal influenza in adults and children-diagnosis, treatment, chemoprophylaxis, and institutional outbreak management:..., Harper [/bib_ref] Obesity, which was not previously recognized as a risk factor for severe influenza, has been observed in a substantial proportion of patients with severe pH1N1 illness. [bib_ref] Hospitalized Patients with 2009 H1N1 Influenza in the United States, Jain [/bib_ref] [bib_ref] Critically ill patients with 2009 influenza A(H1N1) infection in Canada, Kumar [/bib_ref] [bib_ref] Critically Ill Patients With 2009 Influenza A(H1N1) in Mexico, Dominguez-Cherit [/bib_ref] [bib_ref] Factors associated with death or hospitalization due to pandemic 2009 influenza A(H1N1)..., Louie [/bib_ref] Two recent studies comparing hospitalizations and deaths from pH1N1 with the source population suggested that obesity may be independently associated with severe outcomes because of pH1N1. [bib_ref] Risk factors for hospitalisation and poor outcome with pandemic A ⁄ H1N1..., Nguyen-Van-Tam [/bib_ref] , [bib_ref] Morbid obesity as a risk factor for hospitalization and death due to..., Morgan [/bib_ref] We conducted a case-control study to assess risk factors associated with hospitalization and severe outcomes (ICU admission or death) among patients with confirmed pH1N1 illness during the first pandemic wave in the province of Quebec, Canada.
# Methods
## Population
In the province of Quebec, Canada (population 7AE8 million; the second largest provincial population in Canada), all respiratory specimens from patients with suspected pH1N1 infection between April 16th and July 1st, 2009, attending primary care clinics or any hospital were collected and sent to the provincial reference laboratory for testing, along with basic demographic information on patients. Beginning in July, testing protocols changed in the province of Quebec such that only patients attending designated sentinel sites or admitted to hospital were offered laboratory testing for pH1N1 and five additional laboratories were designated to conduct testing. Specimens were tested by reverse transcription polymerase chain reaction (RT-PCR). A positive result by both RT-PCR assays targeting the pH1N1 M and HA genes was required for laboratory confirmation. All hospitalized and non-hospitalized cases with onset of disease during the study period and included in the confirmed case registry of the provincial reference laboratory were considered eligible for the study. A separate provincial registry for patients hospitalized with confirmed pH1N1 infection including name of the hospital and underlying medical conditions was used to validate information related to hospitalization through the admitting hospitals and public health departments of the four regions included in the study. Patients with laboratory-confirmed pH1N1 illness admitted to a hospital for ‡24 hours were eligible as cases. Severe disease was defined as admission to ICU and ⁄ or death. Controls were non-hospitalized patients with laboratory-confirmed pH1N1 illness.
## Study period and data collection
The study period was from May 25th to July 1st, a period that includes approximately 80% of the confirmed cases detected during the first wave of pH1N1 activity in Quebec [fig_ref] Figure 1: Non-hospitalized and hospitalized pH1N1 cases in the province of Quebec during the... [/fig_ref]. During that period, specific diagnostic testing was well established, almost all cases were indigenous, i.e., unrelated to international travel, and laboratory-confirmed case ascertainment was consistent. The study was conducted in the four most affected regions (Montreal, Monteregie, Laval, and Quebec City), which accounted for 78% of all hospitalized and 83% of all non-hospitalized cases in the province of Quebec. Trained interviewers called eligible patients between July 17th and August 10th and collected demographic and clinical data using a standardized questionnaire. For children under 14 years, the respondents were parents. For deceased patients, the respondents were close family relatives. The number of non-hospitalized cases was planned to be proportional to the number of hospitalized cases for each region. For patients with discrepancies in the data, public health departments contacted the relevant hospital and returned corrected validated data to the study team, as well as to the provincial registry.
This study was carried out as an emergency outbreak investigation under the legal mandate of the provincial Chief Medical Officer, and no research ethics board approval was sought.
# Statistical analysis
Proportions were compared using chi-square test or Fisher exact test when appropriate; distributions of age were compared with Wilcoxon rank sum non-parametric test. Odds ratios (OR) to identify risk factors for hospitalization and severe disease were derived by univariate and multivariable unconditional logistic regression. The lowest risk of death was observed in the <5 and in the 5-19 year age groups [fig_ref] Table 1: Participant characteristics by study group AIDS [/fig_ref]. Because the <5 year age group has a high risk for hospitalization and complications for seasonal influenza, we wanted to assess the impact of the pH1N1 virus specifically on this age group. Consequently, the 5-19 age group was selected as the reference category. Underlying medical conditions were analyzed as a group ( ‡1 underlying medical condition) or individually for the most frequent categories (cardiac, pulmonary, diabetes mellitus, asthma, immunosuppression, neurological, and renal conditions). Other underlying conditions (e.g., anemia, cancer, liver, and metabolic diseases) were merged into the 'other conditions' category. Obesity was defined as a body mass index (BMI) ‡30 kg ⁄ m 2 for adults ‡18 years and as a BMI ‡95th percentile for children 2-18 years. [bib_ref] Establishing a standard definition for child overweight and obesity worldwide: international survey, Cole [/bib_ref] In adults, morbid obesity was defined as a BMI ‡40 kg ⁄ m 2 . Indicator variables were created for non-applicable values (e.g., pregnancy for <14 or ‡50 year-old-woman, or man; obesity for <2-year-old; smoking in <14-year-old; health care workers in <20-or >65-year-old); as well as for variables with missing values [fig_ref] Table 1: Participant characteristics by study group AIDS [/fig_ref]. Variables significantly associated with the outcomes of interest in univariate analysis (with a P < 0AE05) or known as potential confounders (gender, seasonal influenza vaccination, antiviral therapy before admission, and pregnancy) were then entered into multivariable analysis. Fully adjusted final models for hospitalization and severe disease included gender, age, education, health care workers (HCW), smoking, seasonal influenza vaccination in 2008-2009, delay in consultation, antiviral use before hospital admission, pregnancy, obesity, and subcategories of underlying medical conditions. The fully adjusted final model for death included gender, age, education, delay in consultation, antiviral use before hospital admission, obesity and sub-categories of underlying medical conditions. Sensitivity analyses were performed by excluding missing data and by excluding HCW from the analysis.
Statistical analysis was performed using SAS software, version 9.1 (SAS Institute Inc., Cary, NC, USA). A two-sided P value of <0AE05 was considered to indicate statistical significance.
# Results
For the 1033 non-hospitalized patients, a phone number was unknown for 300 (29%) and 286 (28%) could not be reached despite at least five separate attempts . Among the 447 persons who were reached, 11 (2%) refused to participate, and one was excluded because of failure to speak English or French. The unreachable patients were comparable to those reached with regards to the proportion of men ⁄ women (44% ⁄ 56% versus 47% ⁄ 53%, respectively, P = 0AE37) and to the age distribution (mean ⁄ median age 26AE3 ⁄ 24 versus 27AE2 ⁄ 22, Wilcoxon test, P = 0AE64).
Among the 435 non-hospitalized patients who agreed to participate, 40 (9%) reported that they had been hospitalized ‡24 hours with their pH1N1 illness. These 40 patients were re-classified accordingly, giving a final sample of 395 non-hospitalized patients . An overall sample of 395 was considered proportional to the number of hospitalized cases and further attempts to reach non-hospitalized patients were stopped at this point.
For the 375 hospitalized cases, the phone number was unknown for 48 (13%) and 25 (7%) could not be reached . Among the 302 hospitalized cases that were reached, 4 (1%) refused to participate, one was excluded because of failure to speak English or French, and 16 (5%) were excluded because they declared reasons other than pH1N1 infection as the primary cause of hospitalization. Data on hospitalized patients were validated with the admitting hospital. The final sample was 321 hospitalized cases (including the 40 re-classified patients), of which 47 (15%) were admitted to an ICU and 15 (5%) died. All deaths occurred in hospital, and four of the deceased patients were not admitted to the ICU before death.
When comparing non-hospitalized and hospitalized cases in univariate analysis, factors significantly associated with hospitalization were as follows: age <5, 50-59, and ‡60 years (compared to age 5-19 years); lower education level; smoking; consulting a physician very early (the day of disease onset) or late (5 days or more after disease onset); having at least one underlying medical condition; and obesity (especially morbid obesity; [fig_ref] Table 2: Odds ratios for hospitalization or ICU admission ⁄ death for various potential... [/fig_ref]. Health care workers represented 17% of non-hospitalized cases but only 2% of hospitalized patients. Antiviral therapy was taken by 22% of non-hospitalized cases and 17% of hospitalized cases before their hospital admission. Asthma was the underlying medical condition most frequently present in all types of cases and increased the risk of being hospitalized by 2AE6-fold. The risk of hospitalization was highest (OR = 24AE8) in patients with neuromuscular disorders. Hospitalized cases declared having anemia (mostly sickle cell anemia) more frequently than non-hospitalized patients (5% versus 0% for anemia; 4% versus 0% for sickle cell anemia, P £ 0AE001 for both comparisons).
In multivariable analysis, age <5 years (OR = 5AE5), consultation ‡5 days after illness onset (OR = 1AE9), and presence of ‡1 underlying chronic condition (OR = 4AE9) were significantly associated with hospitalization [fig_ref] Table 2: Odds ratios for hospitalization or ICU admission ⁄ death for various potential... [/fig_ref]. All underlying chronic conditions, with the exception of immune-suppression, were significantly associated with hospitalization. Health care workers had lower risk of being hospitalized.
Among hospitalized patients, none of the examined factors further significantly increased the risk of admission to an ICU or death. Hospitalized patients who took antiviral medication before hospital admission had significant reduction in the risk of severe disease [OR = 0AE3; 95% confidence interval (CI): 0AE1-0AE8; P = 0AE015] [fig_ref] Table 2: Odds ratios for hospitalization or ICU admission ⁄ death for various potential... [/fig_ref]. In a separate model for death among hospitalized cases (data not shown), the only significant association with death was observed for age ‡60 years (OR = 14AE4, 95% CI: 1AE0-197AE3, P = 0AE046) and immune suppression (OR = 7AE3, 95% CI: 1AE5-35AE0, P = 0AE013).
In sensitivity analyses either restricted to patients without missing data or excluding HCW, results were similar (data not shown).
# Discussion
Our data indicate that young age group (<5 years), underlying chronic conditions and late consultation (5 days or more after symptom onset), was associated with increased risk of hospitalization for pH1N1, whereas age ‡60 years and immune suppression were associated with increased risk of death. As observed globally, age is an important risk factor for pH1N1 hospitalization. Our results show a relatively high proportion of children <5 years of age among hospitalized patients but a small proportion in this age category requiring ICU admission and no deaths. This suggests that physicians may be more inclined to admit young children for less severe disease. In contrast, patients aged ‡60 years were disproportionately represented among deaths despite their small number among non-hospitalized patients. This is consistent with other reports showing that the risk of death in patients with pH1N1 increased with advanced age: older individuals may have had a lower inci-dence but if they became infected they were more likely to progress to severe illness. [bib_ref] Critically ill patients with 2009 influenza A(H1N1) infection in Canada, Kumar [/bib_ref] [bib_ref] Factors associated with death or hospitalization due to pandemic 2009 influenza A(H1N1)..., Louie [/bib_ref] [bib_ref] Infection and death from influenza A H1N1 virus in Mexico: a retrospective..., Echevarria-Zuno [/bib_ref] [bib_ref] Mortality from pandemic A ⁄ H1N1 2009 influenza in England: public health..., Donaldson [/bib_ref] [bib_ref] Risk of severe outcomes among patients admitted to hospital with pandemic (H1N1)..., Campbell [/bib_ref] Presence of underlying chronic conditions was the strongest and most consistent factor contributing to pH1N1 hospitalization, ICU admission, and death. The risk of hospitalization was similar for each specific chronic condition except for immune suppression and neurological disease. Immune suppression was not associated with hospitalization. However, once hospitalized, immune suppression was the only underlying condition associated with death. Neurological diseases as a group showed the strongest association with hospitalization when compared to non-hospitalized patients. Since 2005, both the National Advisory Committee on Immunization in Canada (NACI) and the Advisory Committee on Immunization Practice (ACIP) in the United States have recommended seasonal influenza vaccine for people with cognitive dysfunction, spinal cord injury, seizure disorder, neuromuscular disorders or any such conditions that can compromise respiratory function or the handling of respiratory secretions or that can increase the risk of aspiration. [bib_ref] Prevention and control of Influenza Recommendations of the Advisory Committee on Immunization..., Harper [/bib_ref] Such conditions may be of particular concern for pH1N1 given studies showing greater recognition of receptors in the lung and higher likelihood of primary viral pneumonia associated with that virus. [bib_ref] In vitro and in vivo characterization of new swine-origin H1N1 influenza viruses, Itoh [/bib_ref] [bib_ref] Pathogenesis and transmission of swine-origin 2009 A(H1N1) influenza virus in ferrets, Munster [/bib_ref] [bib_ref] Receptor-binding specificity of pandemic influenza A (H1N1) 2009 virus determined by carbohydrate..., Childs [/bib_ref] In line with other reports, we found that a high proportion of patients admitted with pH1N1 had asthma. [bib_ref] Risk factors for hospitalisation and poor outcome with pandemic A ⁄ H1N1..., Nguyen-Van-Tam [/bib_ref] [bib_ref] Clinical characteristics of paediatric H1N1 admissions in Birmingham, Hackett [/bib_ref] Other reports suggested that patients with pH1N1 illness are more likely to have asthma compared to seasonal influenza A [bib_ref] Risk factors and outcomes among children admitted to hospital with pandemic H1N1..., O'riordan [/bib_ref] The point estimate for the association of at least one underlying medical condition with severe outcomes among hospitalized patients estimated in our study (OR = 1AE5, 95% CI: 0AE8-2AE9) is similar to that reported in another study from Canada by Campbell et al. [bib_ref] Risk of severe outcomes among patients admitted to hospital with pandemic (H1N1)..., Campbell [/bib_ref] among patients admitted to hospital with pH1N1 across the 13 provinces and territories of Canada, including Quebec (OR = 1AE5, 95% CI: 1AE1-2AE1). Campbell et al. relied upon enhanced surveillance reporting by provinces and lacked information on underlying medical conditions and aboriginal status for two of the provinces (including the largest province of Ontario); conversely, we actively queried all participants on a standardized and detailed list of variables of interest, minimizing missing information. The larger CI and absence of statistical significance in our study may be explained by a lack of power because our sample size (62 severe
[formula] (0AE6-1AE2) 1AE2 (0AE7-1AE9) 2AE06 (1AE2-3AE7) 2 AE1 (0AE9-5AE1) Age <5 years 5AE8 (3AE2-10AE4) 5AE5 (2AE1-14AE3) 0 AE5 (0AE2-1AE7) 0AE5 (0AE1-1AE9) 5-19 years Reference Reference Reference Reference 20-34 years 0AE9 (0AE6-1AE4) 2AE6 (0AE5-13AE3) 2AE6 (1AE0-6AE6) 2AE1 (0AE2-24AE2) 35-49 years 1AE5 (1AE0-2AE4) 4AE5 (0AE9-23AE2) 3AE6 (1AE5-8AE4) 1 AE7 (0AE2-18AE3) 50-59 years 1AE7 (1AE0-2AE7) 4 AE3 (0AE8-23AE8) 1AE5 (0AE5-4AE2) 0AE6 (0AE04-7AE5) 60 years and older 3AE7 (2AE0-6AE8) 2 AE2 (0AE7-6AE5) 2AE4 (0AE9-6AE3) 1AE1 (0AE2-7AE2) Education [/formula]
High school not completed
[formula] 2AE5 (1AE5-4AE0) 1 AE3 (0AE7-2AE4) 2AE3 (0AE9-6AE3) 1AE5 (0AE4-4AE9) Non-university certificate ⁄ diploma 1AE5 (1AE0-2AE1) 1 AE2 (0AE7-1AE8) 1AE9 (0AE8-4AE6) 1AE4 (0AE5-4AE1) University degree Reference Reference Reference Reference Health care worker 0AE1 (0AE1-0AE3) 0AE2 (0AE1-0AE4) 2 AE8 (0AE7-11AE7) 4AE3 (0AE8-24AE7) Smoking 1AE9 (1AE2-2AE9) 1 AE3 (0AE7-2AE2) 0AE9 (0AE4-2AE0) 0AE7 (0AE3-1AE7) Seasonal influenza vaccination in 2008-2009 1AE3 (0AE9-1AE8) 1AE0 (0AE6-1AE5) 1AE2 (0AE6-2AE1) 0AE8 (0AE4-1AE8) Consultation, days after symptom onset 0-1 days 1AE5 (1AE0-2AE1) 1 AE5 (0AE9-2AE4) 1AE1 (0AE6-2AE3) 1AE7 (0AE7-3AE9) 2-4 days Reference Reference Reference Reference 5 days or more 1AE8 (1AE2-2AE6) 1AE9 (1AE2-2AE9) 1 AE3 (0AE7-2AE6) 1AE5 (0AE7-3AE4) Antiviral use before hospitalization 0AE9 (0AE6-1AE3) 1AE0 (0AE6-1AE6) 0AE4 (0AE2-1AE1) 0AE3 (0AE1-0AE8) Pregnancy (ref: no, female 14-49 years) 1AE9 (0AE8-4AE9) 2AE4 (0AE8-7AE3) 0AE2 (0AE02-1AE7) 0AE2 (0AE02-1AE8) Underlying medical conditions At least one** 4AE5 (3AE2-6AE3) 4AE9 (3AE2-7AE3) 1 AE6 (0AE9-2AE9) 1AE5 (0AE8-2AE9) Cardiac 5AE6 (2AE8-11AE4) 2AE7 (1AE1-6AE5) 1 AE6 (0AE8-3AE5) 2AE1 (0AE8-6AE1) Diabetes 4AE3 (2AE2-8AE4) 2AE7 (1AE1-6AE5) 1 AE9 (0AE9-4AE0) 1AE4 (0AE5-3AE8) Pulmonary 3AE9 (2AE1-7AE2) 2AE5 (1AE2-5AE4) 1 AE8 (0AE8-3AE7) 1AE7 (0AE8-4AE3) Asthma 2AE6 (1AE6-4AE0) 2AE8 (1AE6-4AE7) 0 AE7 (0AE3-1AE5) 0AE5 (0AE3-1AE3) Immune-suppression 1AE8 (0AE99-3AE4) 0AE8 (0AE4-1AE9) 2AE4 (1AE0-5AE7) 3 AE0 (0AE9-9AE7) Neuromuscular 24AE8 (3AE3-185AE7) 22AE2 (2AE6-186AE0) 1 AE1 (0AE4-3AE5) 0AE9 (0AE2-3AE6) Renal 5AE8 (1AE9-17AE3) 4AE8 (1AE3-17AE7) 0 AE8 (0AE2-3AE0) 0AE4 (0AE08-2AE0) Other 6AE1 (2AE7-14AE1) 5AE6 (2AE0-15AE7) 2 AE1 (0AE9-4AE6) 1AE2 (0AE4-3AE4) Obesity (BMI ‡30) 2AE2 (1AE4-3AE0) 1 AE3 (0AE8-2AE0) 1AE1 (0AE6-2AE0) 1AE0 (0AE5-2AE3) BMI ‡40 (only in adults)*** 3AE1 (1AE2-8AE4) 2 AE0 (0AE6-6AE2) 0AE9 (0AE2-3AE3) 0AE4 (0AE09-2AE0) [/formula]
BMI, body mass index; ICU, intensive care unit. In bold, P < 0AE05. *Adjusted for all variables listed in the table, excepting underlying condition, at least one; and BMI defined as ‡40. **For this variable, a separate model was constructed (multivariable adjusted for age, gender, obesity, being a health care worker, seasonal influenza vaccination, smoking, delay in consultation, antiviral use before admission, pregnancy, and education). ***For this variable, a separate model was constructed only in adults (multivariable adjusted for age, gender, underlying condition, being a health care worker, seasonal influenza vaccination, smoking, delay in consultation, antiviral use before admission, pregnancy, and education).
outcomes and 259 non-severe hospitalized) is smaller than that of Campbell et al. (308 severe and 1171 non-severe). Some differences in associations for severe outcomes with age groups and specific underlying conditions may also be explained by differences in populations (aboriginal status reported for 24% of the patients in Campbell's study, compared to <1% in our study), the more extensive list of adjustment variables we included, differences in the specific study period (April-July versus April-September for Campbell et al.), as well as differences in the propensity for testing and admitting patients. While obesity and morbid obesity were associated with hospitalization in univariate analysis, their effect disappeared after adjustment for gender, age, smoking, education, and underlying medical conditions. Other conditions associated with obesity, such as hypertension, obstructive sleep apnea, or obesity-hypoventilation syndrome, 10 not considered in our study, may also have contributed to this observation. Animal models suggest that obesity may play a specific and independent role in the pathogenesis of seasonal influenza through compromised T-cell-mediated immunity. [bib_ref] Diet-induced obese mice have increased mortality and altered immune responses when infected..., Smith [/bib_ref] [bib_ref] Selective impairment in dendritic cell function and altered antigen-specific CD8+ T-cell responses..., Smith [/bib_ref] Recently, advisory committees in Canada and the United States have added morbid obesity to influenza vaccine recommendations for the 2010-2011 season.Lower education and smoking were also initially found to be associated with hospitalization in univariate analysis in our study, but their effect disappeared in multivariable analysis. Ultimately, underlying medical conditions appear to be the most robust factor associated with hospitalization for pH1N1 illness. Such conditions are known to be more prevalent among the socially and economically disadvantaged segments of the population which we did not specifically assess. [bib_ref] Disease and disadvantage in the United States and in England, Banks [/bib_ref] The higher risk for hospitalization when seeking medical care late in the course of illness may be related to influenza complications such as advanced acute respiratory distress syndrome or bacterial superinfection. Bacterial superinfection was observed in 4-20% of patients hospitalized with pH1N1 infection, 13-29% of fatal pH1N1 cases, 5,10,20 and 43% of fatal pH1N1 pediatric cases.At least one additional pathogenic respiratory virus or bacterium was found in nasopharyngeal aspirates of up to 76% of medically attended confirmed pH1N1 infection; however, only Streptococcus pneumoniae was associated with severe pH1N1 disease. [bib_ref] Streptococcus pneumoniae coinfection is correlated with the severity of H1N1 pandemic influenza, Palacios [/bib_ref] While we have not been able to assess antiviral therapy administered during the hospital stay, our results suggest that antiviral use before hospital admission may protect against severe pH1N1 outcomes and are consistent with data from other studies for seasonal [bib_ref] Antiviral therapy and outcomes of influenza requiring hospitalization in Ontario, Mcgeer [/bib_ref] [bib_ref] Factors associated with early hospital discharge of adult influenza patients, Lee [/bib_ref] and pH1N1 influenza. [bib_ref] Hospitalized Patients with 2009 H1N1 Influenza in the United States, Jain [/bib_ref] [bib_ref] Correlates of severe disease in patients with 2009 pandemic influenza (H1N1) virus..., Zarychanski [/bib_ref] [bib_ref] Risk factors for hospitalisation and poor outcome with pandemic A ⁄ H1N1..., Nguyen-Van-Tam [/bib_ref] [bib_ref] Clinical characteristics of paediatric H1N1 admissions in Birmingham, Hackett [/bib_ref] The proportion of HCWs in the Quebec population is 3AE6% compared to 17% observed among non-hospitalized patients enrolled in our study. This is unlikely to be explained by a greater risk of pH1N1 among HCWs but more probably reflects the easier access to influenza testing for HCWs compared to the rest of the population. This latter hypothesis is supported by the 5-fold lower risk of hospitalization and severe disease for HCWs compared to non-HCW. The oversampling of HCW did not appear to influence our results because findings from multivariable analyses were robust with adjustment for HCW status and analyses excluding HCW showed similar estimates.
The findings we report are subject to several limitations. First, we reached the majority of hospitalized cases, but only 43% of non-hospitalized cases. We attained our enrollment target and few people contacted refused participation. Patients that were reached were comparable to those not reached on the basis of sex and age distribution, but it is possible that individuals who were not reached had other characteristics that differed. Second, we cannot exclude a role for recall bias: hospitalized cases may have better memory for events related to their hospitalization. Similarly, proxy responses provided by close relatives of deceased patients may differ from the rest of our study population. Social stigma may have lead to underreporting of weight, smoking, or lower education. This is likely to apply equally to non-hospitalized and hospitalized cases which would then decrease the strength of associations found. Third, for non-hospitalized cases, testing for pH1N1 may have been subject to different health care-seeking behaviors by patients or different levels of clinical concern by physicians. In our analyses, we adjusted for being a HCW, as well as for education, smoking, and seasonal influenza vaccination which may be markers of health-seeking behavior. However, we cannot exclude the role of residual bias unaccounted for by our analyses. Physician vigilance may have led to persons with underlying conditions -hospitalized or non-hospitalized -being more often tested (and detected as pH1N1 positive) compared to those without underlying conditions. This would also have led to an underestimate of the effect of chronic conditions on the risk of hospitalization and disease severity. We identified age <5 years as a risk factor for hospitalization but sample size precluded further age stratification within that young age group. Infants and toddlers <24 months of age have been at recognized increased risk for hospitalization because of seasonal influenza for several years, but preschool children 2-5 years have not previously been identified at higher risk for hospitalization because of seasonal influenza, although they experience more medical visits.Thus, the more precise age-specific risk of pH1N1 hospitalization within this young age group warrants closer examination. Finally, while this study was well powered to assess the risk factors for hospitalizations, fewer participants admitted to ICU or dying were represented, limiting Risk factors for pandemic H1N1 influenza ª 2011 Blackwell Publishing Ltd, Influenza and Other Respiratory Viruses, 5, 247-255 the power to assess factors associated with those most severe outcomes.
In summary, we identified young age, underlying conditions and delayed consultation 5 days or more after illness onset as factors associated with higher risk of pH1N1 hospitalization. Older age ( ‡60 years) and immune suppression were associated with death among hospitalized patients. Antiviral medication taken before hospital admission did not protect against hospitalization but protected against subsequent more severe outcomes of ICU admission or death. Most variables associated with hospitalization and severe pH1N1 disease are also recognized risk factors for severe seasonal influenza. Further research is needed to better define whether obesity per se has an independent role in increasing the risk for severe pH1N1 illness.
[fig] Figure 1: Non-hospitalized and hospitalized pH1N1 cases in the province of Quebec during the period of April 16-August 30, 2009. [/fig]
[table] Table 1: Participant characteristics by study group AIDS; chemo-and radiotherapy for cancer, long-term corticosteroid treatment. Confirmed by a pulmonary function test or severe enough to have required hospitalization or systemic corticosteroids in the past 12 months.Mostly chronic obstructive pulmonary disease and congenital pulmonary disorders in children. àà Mostly cerebral palsy, muscular dystrophy, developmental delay.In children defined according to Ref.25. § Defined only in adults ‡18 years. [/table]
[table] Table 2: Odds ratios for hospitalization or ICU admission ⁄ death for various potential risk factors [/table]
|
The Assessment of Selected Bone and Cartilage Biomarkers in Psoriatic Patients from Poland
Background. Psoriasis is an inflammatory disease in which joints involvement may be insidious and difficult to detect. Bone and cartilage biomarkers may be helpful in screening patients with psoriasis for psoriatic arthritis (PsA). Objectives. To assess bone and cartilage serum biomarkers in psoriasis. Methods. The study was conducted in 2014 and included 61 psoriatic patients and 30 healthy individuals. In both groups, the serum concentrations of soluble receptor activator of nuclear factor-B ligand (sRANKL), cartilage oligomeric matrix protein (COMP), osteoprotegerin (OPG), and interleukin-20 (IL-20) were examined. Severity of skin lesions was assessed by Psoriasis Area and Severity Index (PASI), body surface area (BSA), and Physician Global Assessment (PGA) scores. Results. The duration of psoriasis was from 1 year to 45 years. 22 patients suffered from concomitant PsA. The mean value of PASI was 23.1 ± 12.0 and BSA was 27.6 ± 20.6%. COMP, OPG, and IL-20 concentrations in psoriatic patients were significantly higher than in the control group. OPG/sRANKL ratio was significantly lower in PsA patients than in psoriatic patients without arthritis. Conclusions. Results of the conducted study suggest that COMP, OPG, IL-20, and OPG/sRANKL ratio may appear useful biomarkers of bone and cartilage involvement in psoriasis.
# Introduction
Psoriatic arthritis is a clinical manifestation of psoriasis affecting the musculoskeletal system, which according to different sources is observed in 8-10% to 30% of patients with psoriasis [bib_ref] Release of cartilage and bone macromolecules into synovial fluid: differences between psoriatic..., Månsson [/bib_ref] [bib_ref] Prevalence of joint disease in patients with psoriasis: implications for therapy, Zachariae [/bib_ref]. However, discrete subclinical changes in the joints and entheses may occur before the onset of PsA and may be more frequent in patients with psoriatic skin disease [bib_ref] Analysis of periarticular bone changes in patients with cutaneous psoriasis without associated..., Simon [/bib_ref]. The PsA development may be insidious with pathological changes induced in the bones, cartilages, and tendons, which are difficult to detect since currently applied diagnostic laboratory and imaging methods are capable of recognizing only already formed abnormalities. Enthesopathy and new bone formation as well as bone loss and cartilage apposition are common in psoriatic patients [bib_ref] Analysis of periarticular bone changes in patients with cutaneous psoriasis without associated..., Simon [/bib_ref] [bib_ref] Value of entheseal ultrasonography and serum cartilage oligomeric matrix protein in the..., Farouk [/bib_ref] [bib_ref] Biomarkers and prognostic stratification in psoriatic arthritis, Bogliolo [/bib_ref]. Bones are known to be metabolically active tissues and two types of cells, osteoblasts and osteoclasts, are involved in their remodeling [bib_ref] Adipokines in psoriatic arthritis patients: the correlations with osteoclast precursors and bone..., Xue [/bib_ref]. Local or systemic bone loss is typical for PsA leading to an increased risk of osteoporotic fractures. The increased bone resorption results from chronic inflammation and side effects induced by prolonged administration of such drugs as methotrexate and corticosteroids as well as reduced physical activity because of pain, oedema, and joint dysfunction [bib_ref] Tumour necrosis factor-related apoptosis-inducing ligand and osteoprotegerin serum levels in psoriatic arthritis, Hofbauer [/bib_ref]. Proinflammatory cytokines released in psoriasis stimulate formation and activation of osteoclasts which are directly responsible for the bone tissue loss [bib_ref] Assessment of osteoporosis in psoriasis with and without arthritis: correlation with disease..., Attia [/bib_ref]. They also stimulate chondrocytes to produce destructive proteases, which leads to proteoglycan loss, destruction of collagen bundles, and COMP release [bib_ref] Value of entheseal ultrasonography and serum cartilage oligomeric matrix protein in the..., Farouk [/bib_ref]. COMP, a homopentamer glycoprotein of the thrombospondin family, is one of the components of extracellular matrix of the joint cartilage. An elevated COMP level in the synovial fluid and serum is thought to reflect the remodeling and repair processes in the cartilage tissue [bib_ref] Release of cartilage and bone macromolecules into synovial fluid: differences between psoriatic..., Månsson [/bib_ref]. Bone modeling at the entheses (enthesiophyte formation) may also occur in the absence of inflammation, but it can be the response to biomechanical stress and repeated microtraumatic injury. It can be recognized as the deep Koebner Phenomenon at the entheseal sites of psoriatic patients [bib_ref] Analysis of periarticular bone changes in patients with cutaneous psoriasis without associated..., Simon [/bib_ref].
## Mediators of inflammation
In PsA bone resorption, the imbalance of the RANK (receptor-activator of NF-B)/RANKL/OPG axis is thought to play a significant role. MCSF (macrophage colonystimulating factor) and RANKL, a ligand for RANK on osteoclasts and their precursors (OCPs), are necessary for the osteoclasts formation from monocytic precursors in the synovial tissue [bib_ref] Circulating mediators of bone remodeling in psoriatic arthritis: implications for disordered osteoclastogenesis..., Dalbeth [/bib_ref]. RANK and its coreceptor, RANKL, are members of the TNF (tumor necrosis factor) superfamily. Synovial fibroblasts and activated T cells are the main source of MCSF and RANKL [bib_ref] Joint remodelling in inflammatory disease, Schett [/bib_ref]. RANKL, a membrane protein on the osteoblasts, interacts with RANK, a type I transmembrane receptor on the bone marrow macrophages inducing their differentiation into osteoclasts [bib_ref] Adipokines in psoriatic arthritis patients: the correlations with osteoclast precursors and bone..., Xue [/bib_ref]. It has been shown that RANKL plays a significant role in regulation of numerous functions of the dendritic cells (DCs), such as enhancement of T cell activation capacity, increased release of proinflammatory cytokines, and prolonged cell survival. Moreover, RANKL is expressed on inflamed or activated keratinocytes and it induces activation of epidermal Langerhans cells (LCs) [bib_ref] Tissue microarray analysis of RANKL in cutaneous lupus erythematosus and psoriasis, Toberer [/bib_ref].
Numerous proinflammatory cytokines, including TNF, IL-1beta, IL-6, IL-15, IL-17, and IL-23, exhibit synergic activity with RANKL in proliferation and differentiation of osteoclasts [bib_ref] Assessment of osteoporosis in psoriasis with and without arthritis: correlation with disease..., Attia [/bib_ref] [bib_ref] Anti-IL-20 monoclonal antibody inhibits the differentiation of osteoclasts and protects against osteoporotic..., Hsu [/bib_ref].
Similar to MCSF and RANKL, IL-20 stimulates osteoclast differentiation. Hsu et al. [bib_ref] Anti-IL-20 monoclonal antibody inhibits the differentiation of osteoclasts and protects against osteoporotic..., Hsu [/bib_ref] have shown that IL-20 increases the RANK expression on the osteoclast precursors from the bone marrow and it plays a significant role in their differentiation. Furthermore, IL-20 contributes to RANKL expression on osteoblasts.
Apart from RANKL, osteoprotegerin, a glycoprotein of the TNF family, is another receptor for RANK, which acts as a decoy receptor thereby, preventing RANKL to RANK binding and inhibiting the osteoclastogenic process. OPG is also regarded as a marker of periostitis and new bone formation [bib_ref] Biomarkers and prognostic stratification in psoriatic arthritis, Bogliolo [/bib_ref]. The OPG/RANKL ratio is widely used to determine the bone remodeling and osteoclastogenesis [bib_ref] Adipokines in psoriatic arthritis patients: the correlations with osteoclast precursors and bone..., Xue [/bib_ref].
Determination and monitoring of biomarkers of the bone and cartilage turnover seem to be a promising concept [bib_ref] Value of entheseal ultrasonography and serum cartilage oligomeric matrix protein in the..., Farouk [/bib_ref] [bib_ref] Biomarkers in psoriatic arthritis: recent progress, Chandran [/bib_ref]. Since the metabolites which are formed in the process of remodeling and destruction are not only present in the synovial fluid but also released into the bloodstream, they may turn out to be useful biomarkers of the bone and cartilage changes.
The aim of the study was to evaluate concentrations of selected biomarkers involved in the bone and cartilage remodeling in the serum of psoriatic patients.
# Materials and methods
## Studied group.
The study was conducted in 2014 in patients hospitalized in the Department of Dermatology, Venereology, and Pediatric Dermatology, Medical University of Lublin, Poland, because of psoriasis exacerbation.
The study comprised 61 male psoriatic patients and 30 male healthy controls. Inclusion criteria were the duration of psoriasis for at least one year, presence of active psoriatic skin lesions on the skin, and age above 18 years old. Patients who received any systemic treatment for psoriasis within last 12 months or applied steroids, retinoids, anthralin, or vitamin D analogs on the skin within last 2 months were excluded from the study. Patients with presence of cancer, hematologic, hepatic or renal disorder, and history of musculoskeletal diseases other than PsA were excluded from the study. Twenty-two patients (36.07%) suffering from PsA met the Classification of Psoriatic Arthritis (CASPAR) criteria for PsA. All the PsA patients had polyarticular, asymmetrical subset of the disease (more than 5 joints were affected). Demographic data, medical history, blood, and serum for assessment of the selected biomarkers were collected from all the participants.
## Assessment of psoriasis severity.
The skin changes severity was assessed using Psoriasis Area and Severity Index (PASI), Body Surface Area (BSA), and Physician Global Assessment (PGA) scores. PASI is the most widely used tool for the measurement of severity of psoriasis. It scores the average redness, thickness, and scaling of the lesions (0-4 scale) and percentage of affected area. PASI scores range from 0 (no active disease) to 72 (the most severe disease) [bib_ref] Biomarkers in psoriatic arthritis: recent progress, Chandran [/bib_ref]. BSA is the percentage of Body Surface Area involved in disease. PGA is another frequently used scale, rated on a scale from clear or inactive to severe or most active [bib_ref] Biomarkers in psoriatic arthritis: recent progress, Chandran [/bib_ref] [bib_ref] Psoriasis assessment tools in clinical trials, Feldman [/bib_ref]. It is more subjective than PASI, but it is an easier and faster method to use in clinical practice [bib_ref] A quantitative definition of severe psoriasis for use in clinical trials, Feldman [/bib_ref].
## Assessment of cytokine serum concentrations in the psoriatic patients and controls.
Blood samples were collected from psoriatic patients and controls and were centrifuged for 15 minutes at 1000 ×g. Then, serum samples were subdivided into small aliquots to be stored at −80 ∘ C until tested for biomarkers levels. In the studied psoriatic patients as well as the control group, the concentrations of human sRANKL (total) ELISA (BioVendor, catalog number RD193004200R), human COMP (R&D Systems, catalog number DCMP0), human OPG (Demeditec Diagnostics GmbH, catalog number DE1940032), and human IL-20 (R&D Systems, catalog number DL200) were determined, according to the manufacturer's instructions. The minimum detectable dose (MDD) of sRANKL was determined to be 0.4 pmol/L. The coefficient of variation (CV) for sRANKL was 7.25-11.51% for intra-assay precision and 11.21-12.77 for interassay precision. MDD of COMP was 0.005 ng/mL. CV for COMP was 2.1-3.8% for intra-assay precision and 4.3-4.8 for interassay precision. MDD of OPG was determined to be 0.1 pmol/L. CV for OPG was 2.5-4.9 for intra-assay precision and 1.7-9.0 for interassay precision. MDD of IL-20 was 2.63 pg/mL. CV for IL-20 was 6.2-9.5 for intra-assay precision and 4.8-9.0 for interassay precision.
ESR and CPR were performed in the hospital central laboratory.
The study was approved by the Polish Local Ethics Committee.
## Statistical analyses.
First we analysed sociodemographic and clinical data of psoriatic patients (psoriasis with and without arthritis combined, and separately psoriasis without arthritis and PsA) and control group. We used -test to Then, we fitted logistic regression models to explain which biomarkers have association with psoriasis and PsA. We estimated both univariate and multivariate regression models to identify biomarkers associated with psoriasis (psoriasis with and without arthritis combined) versus control group. We also fitted univariate regression models to identify biomarkers associated with PsA versus psoriasis without arthritis. Because only one biomarker was associated with PsA versus psoriasis without arthritis, there was no estimated multivariate logistic regression model. Multiple comparisons of biomarkers between the three groups (psoriasis without arthritis, PsA, and control group) were performed using polychotomous logistic regression models. The reference category was set as the control group, so we compared the effects of biomarkers in psoriatic patients without arthritis versus the control group and in PsA patients versus the control group.
As results of logistic regression analyses, we presented odds ratio (OR) that indicates change per unit increase in biomarker concentration and CI (confidence intervals) for OR. We constructed receiver operating characteristic (ROC) curves; the areas under the ROC curve were estimated to evaluate how accurately regression models classify patients into analysed groups. The cut-off values of biomarkers for diagnosis of disease were estimated as the point of intersection of curves in which both sensitivity and specificity of patients' classification for logistic regression model were the highest.
Mean values ( ) and standard deviations (SD) were calculated for continuous variables, or absolute numbers ( ) and relative numbers (%) of occurrence of items of categorical variables.
The Pearson correlation coefficient ( ) was used to investigate mutual correlations between the biomarkers as well as correlation between biomarkers and clinical data in psoriatic patients.
Statistical analysis was performed using SAS System software.
In all statistical tests, significance level was set at 0.05.
# Results
Sociodemographic characteristics and clinical data of psoriatic patients and control group are presented in [fig_ref] Table 1: Sociodemographic characteristics and clinical data of psoriasis patients and control group [/fig_ref]. The studied psoriatic patients' age was 46.4 ± 14.1 years on average and did not significantly differ with respect to control group's age ( = 0.654; = 0.516). Psoriatic patients with and without arthritis did not differ significantly with respect to age ( = 0.195; = 0.846). Duration of psoriasis ranged from 1 to 45 years, 20.8 ± 12.0 years on average. Duration of PsA ranged from 1 to 25 years, 10.1 ± 6.2 years on average. 21 patients (34.43%) had positive psoriatic family history.
The PASI value in the studied group ranged from 7.5 to 56.9, 23. [fig_ref] Table 2: The biomarkers serum concentrations in the psoriatic patients compared to the control... [/fig_ref] presents mean concentrations of the biomarkers in psoriatic patients and in the control group and the results of univariate and multivariate logistic regression analysis of psoriasis' probability. Serum concentrations of COMP, OPG, and IL-20 in the psoriatic patients were significantly higher than in the [fig_ref] Figure 1: ROC curve for the reduced multivariate logistic regression model comparing patients with... [/fig_ref]. No significant differences were found in the sRANKL and OPG/sRANKL ratio concentrations between the psoriatic patients and the control group.
## Comparing selected biomarkers' serum concentrations in the studied psoriatic patients to the control group.
ESR and CPR were within the normal ranges and there were no significant differences between the psoriatic patients and the control group as well as between psoriatic and PsA patients. [fig_ref] Table 3: The biomarkers serum concentrations in the psoriatic patients with and without arthritis [/fig_ref] presents mean concentrations of the biomarkers in the psoriatic patients without arthritis and in the PsA patients and the results of univariate logistic regression analysis of psoriasis arthritis probability. No significant differences were found in the sRANKL, COMP, OPG, and IL-20 concentrations between the psoriatic patients without arthritis and PsA patients. OPG/sRANKL ratio was significantly lower in the PsA patients than in psoriatic patients without arthritis. Estimated odds of having psoriatic arthritis decreased by about 8% if OPG/sRANKL ratio increased by about 1. The cut-off values OPG/sRANKL were estimated as 5.533 and less for diagnosis of psoriatic arthritis. It is the point of intersection of curves in which both sensitivity and specificity of patients' classification for logistic regression model were the highest. [fig_ref] Table 4: Logistic regression analyses of the biomarkers serum concentrations in the psoriasis without... [/fig_ref] psoriasis without arthritis increased if serum concentrations of these two biomarkers increased. Serum concentration of COMP was significantly higher in psoriatic patients than in the control group ( = 0.007 and OR > 0), but it was not significantly different between PsA patients and the control group ( = 0.332).
## Comparing selected biomarkers' serum concentrations in the psoriatic patients without arthritis to the psa patients.
## Comparing selected biomarkers' serum concentrations in the psoriatic patients without arthritis and psa patients to the control group.
## Analysis of correlations between the determined biomarkers' concentrations in the studied psoriatic patients.
The OPG concentration was negatively correlated to psoriasis severity measured by the sPGA ( = −0.286; = 0.031) and positively to psoriasis duration ( = 0.337; = 0.010). It means that the lower the sPGA value and the longer the psoriasis duration were, the higher the OPG concentration was. IL-20 concentration was positively correlated to psoriasis severity expressed by PASI ( = 0.418; = 0.001), BSA ( = 0.579; < 0.001), and PGA ( = 0.392; = 0.002). The higher the IL-20 concentration was, the more severe the psoriasis was on average.
No significant correlations were found between the concentrations of the other biomarkers' serum concentration and psoriasis duration, psoriatic arthritis duration, and psoriasis severity expressed by PASI, BSA, and PGA [fig_ref] Table 5: Correlation coefficients between the biomarkers serum concentrations and clinical data in the... [/fig_ref].
In addition, mutual correlations between the concentrations of selected biomarkers in the psoriatic patients were analysed [fig_ref] Table 6: Mutual correlation coefficients between the biomarkers serum concentrations in the psoriatic patients [/fig_ref]. A significant negative correlation between sRANKL and COMP concentrations was found ( = −0.308; = 0.044). An increase in the sRANKL was accompanied by an average decrease in the COMP concentration. No significant correlations were found between the serum concentrations of IL-20 and sRANKL, COMP, and OPG as well as between OPG and sRANKL and COMP.
# Discussion
Gladman et al. [bib_ref] Clinical indicators of progression in psoriatic arthritis: multivariate relative risk model, Gladman [/bib_ref] have observed that 16% of the patients with a 14-year-long psoriasis history developed deformations 6 Mediators of Inflammation in more than five joints. Early diagnosis of PsA (i.e., before its clinical manifestations or irreversible changes have developed) is essential since prompt medical intervention is likely to stop the progress of the disease and prevent joint destruction. Therefore, sensitive and specific diagnostic tools are still in demand. In the light of recent studies, patients with isolated cutaneous psoriasis may have subclinical periarticular bone changes in a form of new bone formation in the entheseal regions of the joints (enthesiophytes) [bib_ref] Analysis of periarticular bone changes in patients with cutaneous psoriasis without associated..., Simon [/bib_ref].
In our own study significantly higher COMP, OPG, and IL-20 serum concentrations in psoriatic patients were found in comparison to the control group; the results were independent of the presence of PsA. This could imply that, in some of the studied psoriatic patients without arthritis, asymptomatic osteoarticular changes are being formed. Hence, in order to prevent formation of irreversible changes and slow down the progress of the disease, close monitoring of such patients and early intervention should be executed.
In the cartilage, made up of chondrocytes imbedded in the matrix, proteoglycans and type II collagen are the fundamental constituents. Activation of cytokines (IL-1, IL-17, TNF, and oncostatins) stimulates chondrocytes to release destructive proteases, which will result in the proteoglycan loss and destruction of collagen bundles. Consequently, cartilage macromolecules, such as aggrecan and COMP, are released into the synovial fluid and serum, which may be reflective of the cartilage turnover [bib_ref] Release of cartilage and bone macromolecules into synovial fluid: differences between psoriatic..., Månsson [/bib_ref].
The studies conducted so far indicated that in patients with PsA the level of COMP in the synovial fluid and serum is increased [bib_ref] Release of cartilage and bone macromolecules into synovial fluid: differences between psoriatic..., Månsson [/bib_ref] [bib_ref] Increased serum levels of cartilage oligomeric matrix protein in patients with psoriasis..., Skoumal [/bib_ref]. COMP expression is confirmed not only in the cartilage but also in tendons, entheses, and ligaments as well. Therefore, COMP expression may be regarded as an early marker of cartilage destruction and turnover in the PsA patients [bib_ref] Value of entheseal ultrasonography and serum cartilage oligomeric matrix protein in the..., Farouk [/bib_ref].
Månsson et al. [bib_ref] Release of cartilage and bone macromolecules into synovial fluid: differences between psoriatic..., Månsson [/bib_ref] have shown that COMP concentration in the synovial fluid is significantly higher in PsA than rheumatoid arthritis (RA). This, according to the authors, may be related to the degradation and release of newly synthesised molecules reflecting a repair process. Furthermore, this repair process appears to be sufficient for preventing permanent cartilage destruction [bib_ref] Release of cartilage and bone macromolecules into synovial fluid: differences between psoriatic..., Månsson [/bib_ref].
The study results concerning correlation of COMP and PsA activity are inconclusive. According to some researchers, serum COMP correlates with PsA activity as well as with inflammation indicators such as CRP, ESR (erythrocyte sedimentation rate), and the number of oedemic joints [bib_ref] Increased serum levels of cartilage oligomeric matrix protein in patients with psoriasis..., Skoumal [/bib_ref]. Thus, serum COMP could reflect the PsA activity. However, Månsson et al. [bib_ref] Release of cartilage and bone macromolecules into synovial fluid: differences between psoriatic..., Månsson [/bib_ref] , similar to our findings, did not reveal any correlation between COMP concentration and levels of inflammatory indicators.
There are reports which present COMP as a biomarker in psoriatic patients without arthritis. Skoumal et al. [bib_ref] Increased serum levels of cartilage oligomeric matrix protein in patients with psoriasis..., Skoumal [/bib_ref] found out that the severity of skin lesions had no effect on the level of serum COMP which depended only on the coexisting inflammatory changes in the joints. Their patients with active PsA presented markedly elevated serum COMP levels in comparison to the PsA patients with low clinical and laboratory activity of the disease. The results of ultrasound (US) examination in a group of 60 psoriatic patients (30 with psoriasis without arthritis and 30 with PsA) revealed that the prevalence of entheseal abnormalities was not significantly different between psoriatic patients without arthritis and PsA patients ( = 0.19) [bib_ref] Value of entheseal ultrasonography and serum cartilage oligomeric matrix protein in the..., Farouk [/bib_ref]. Therefore, according to Farouk et al. [bib_ref] Value of entheseal ultrasonography and serum cartilage oligomeric matrix protein in the..., Farouk [/bib_ref] , the level of sCOMP is a more sensitive biomarker, whereas the ultrasound is more specific. In both studied groups, a correlation was found between COMP and CRP as well as between COMP and DAS28. However, no correlation with PASI was detected. On the basis of performed US, the authors concluded that enthesitis may be asymptomatic and some cases of Achilles tendon enthesopathy remain undetected/undiagnosed. The elevated COMP levels in the psoriatic patients without arthritis and PsA patients suggest that undetected articular involvement may also be present in psoriatic patients without arthritis [bib_ref] Value of entheseal ultrasonography and serum cartilage oligomeric matrix protein in the..., Farouk [/bib_ref]. All these findings give grounds for close monitoring of psoriatic patients without arthritis with elevated sCOMP levels so that further PsA development could be prevented.
The RANK-RANKL-OPG axis may be another group of useful biomarkers in psoriasis. OPG, also known as the osteoclast differentiation-inhibiting protein, is an important osteoclastogenesis inhibitor. The studies conducted so far confirm the role of RANKL-promoted osteoclastogenesis in the PsA bone resorption and osteoporotic fractures [bib_ref] Adipokines in psoriatic arthritis patients: the correlations with osteoclast precursors and bone..., Xue [/bib_ref]. The increased risk of developing osteoporosis by psoriatic patients is largely connected with an ongoing chronic inflammatory process [bib_ref] Assessment of osteoporosis in psoriasis with and without arthritis: correlation with disease..., Attia [/bib_ref]. In our study, OPG concentration was significantly higher in psoriatic than control group and was positively correlated with psoriasis duration; the longer the disease duration, the higher the OPG concentration. However, the negative correlation among OPG concentration and sPGA was noted.
The pathogenetic correlation between the immune system and bone metabolism is well established and OPG and RANKL are known to play a role in the activation of T cells, chronic inflammation, and bone resorption. Activated T cells release RANKL which favours the pathogenesis of inflammatory bone diseases and bone loss. The cytokines released in the inflammatory process stimulate production and activation of the osteoclasts, whereas OPG, as a decoy receptor, increases bone density [bib_ref] Tumour necrosis factor-related apoptosis-inducing ligand and osteoprotegerin serum levels in psoriatic arthritis, Hofbauer [/bib_ref].
In Hofbauer et al. [bib_ref] Tumour necrosis factor-related apoptosis-inducing ligand and osteoprotegerin serum levels in psoriatic arthritis, Hofbauer [/bib_ref] 's study, the serum OPG level was not significantly different between the PsA patients and the control group. Interestingly, PsA females had higher OPG concentrations in comparison with the male patients, which may be explained by the stimulatory effect of estrogens on OPG. The authors' opinion was that low osteoporosis prevalence in PsA women may be explained by the compensatory mechanism of the elevated OPG level [bib_ref] Tumour necrosis factor-related apoptosis-inducing ligand and osteoprotegerin serum levels in psoriatic arthritis, Hofbauer [/bib_ref].
Unfortunately, the positive effect of the elevated OPG in psoriasis is not sufficient to prevent further osteopathogenic processes. Ritchlin et al. [bib_ref] Mechanisms of TNF--and RANKL-mediated osteoclastogenesis and bone resorption in psoriatic arthritis, Ritchlin [/bib_ref] have found that osteoclast precursors are markedly increased in the serum of PsA patients, especially in those with bone erosions confirmed by the radiograph. Moreover, in their immunohistochemical examinations, the authors confirmed RANKL expression on the synovial lining cells, while OPG expression was limited to the endothelium. Because of this, OCPs with RANK receptor are likely to induce osteoclastogenesis in the synovial environment, where high RANKL/OPG ratio is present. The RANKL-RANK interaction results in OCPs differentiation into multinuclear bone-resorbing osteoclasts. Numerous cytokines, including the IL-23/IL-17 axis, increase RANK expression. Also IL-20, released by monocytes, keratinocytes, and Th17 cells and IL-20R in the synovium of PsA patients, increases RANK and RANKL expression [bib_ref] The IL-23/IL-17 axis in psoriatic arthritis, Suzuki [/bib_ref]. In our study, IL-20 was significantly higher in both psoriatic patients without arthritis and PsA patients in comparison with the control group. However, no difference in IL-20 concentration was found among psoriatic patients without arthritis and PsA patients. Hsu et al. [bib_ref] Anti-IL-20 monoclonal antibody inhibits the differentiation of osteoclasts and protects against osteoporotic..., Hsu [/bib_ref] have pointed out that IL-20 is higher in patients with bone loss resulting from osteoporosis. IL-20 stimulates differentiation of osteoclasts; therefore, blocking of this cytokine by monoclonal antibodies may be of therapeutic significance. M-CSF and RANKL are known to be indispensible and sufficient for osteoclasts differentiation. The authors have shown that IL-20 increased RANK expression on the osteoclasts precursors from bone marrow. In addition, IL-20 also increased RANKL expression on osteoblasts. TNF also promoted the formation of osteoclasts, and infliximab application prevented bone loss in PsA patients [bib_ref] Tumour necrosis factor-related apoptosis-inducing ligand and osteoprotegerin serum levels in psoriatic arthritis, Hofbauer [/bib_ref].
Toberer et al. [bib_ref] Tissue microarray analysis of RANKL in cutaneous lupus erythematosus and psoriasis, Toberer [/bib_ref] found a significantly increased RANKL expression in psoriatic epidermis, while it was not observed in the patients with cutaneous lupus erythematosus. This could be explained by Treg functional insufficiency in psoriatic skin lesions that resulted in decreased effector T cell inhibition.
Previous reports suggest that RANKL may be helpful in identification of patients with progressive arthritis or in those with an aggravated course of the disease [bib_ref] Biomarkers and prognostic stratification in psoriatic arthritis, Bogliolo [/bib_ref]. In our study, maybe because of a less severe course of PsA (ESR and CPR were within the normal ranges), sRANKL was not significantly different among psoriatic patients in comparison with the control group.
Attia et al. [bib_ref] Assessment of osteoporosis in psoriasis with and without arthritis: correlation with disease..., Attia [/bib_ref] observed significantly increased OPG serum levels in both psoriatic patients without arthritis and PsA patients suggestive of the presence of osteoporosis regardless of the sex, age, BSA, and PASI. They also found a correlation between the number of affected joints and an increased risk of developing osteoporosis. Therefore, apart from determination of the concentration of individual biomarkers, it is essential that the OPG/RANKL ratio should be calculated. Xue et al. [bib_ref] Adipokines in psoriatic arthritis patients: the correlations with osteoclast precursors and bone..., Xue [/bib_ref] , who made an attempt to determine selected osteoclastogenesis biomarkers, revealed that the concentrations of TNF-alfa, RANKL, and osteoclast precursors were significantly higher in PsA patients than in those with psoriatic patients without arthritis and the control group. The OPG/RANKL ratio was significantly lower in PsA than in psoriatic patients without arthritis. No statistically significant difference in the OPG concentrations between the controls and PsA patients was found, and neither was it found between the patients with erosive and nonerosive PsA. They also found a positive correlation between RANKL and osteoclast precursors' concentrations as well as PsAJAI (Psoriatic Arthritis Joint Activity Index), which further confirms the bone remodeling and RANKL/osteoclastogenesis activity in PsA [bib_ref] Adipokines in psoriatic arthritis patients: the correlations with osteoclast precursors and bone..., Xue [/bib_ref]. Having observed significantly higher values of the OPG/RANKL ratio in psoriatic patients without arthritis in comparison with the controls, the same authors suspected the presence of unidentified factors capable of preventing osteoclastogenesis in psoriatic patients without arthritis. Interestingly, the study results presented by Xue et al. [bib_ref] Adipokines in psoriatic arthritis patients: the correlations with osteoclast precursors and bone..., Xue [/bib_ref] are consistent with the results obtained in our study, where the OPG/sRANKL ratio was significantly higher in the psoriatic patients without arthritis than in PsA patients ( = 0.013). However, the fact that we have determined the levels of biomarkers merely in the serum of psoriatic patients may be a limitation of our study and, therefore, further studies are needed to be performed in this area, for example, simultaneous determination of the cartilage and bone biomarkers in the synovial fluid and psoriatic skin specimens.
It seems likely that determination of the cartilage and bone remodeling biomarkers in psoriatic patients will make it possible to identify the patients with an increased risk of developing PsA. Determination of more than one biomarker appears to increase the possibility of assessing not only severity but also activity of the disease and identification of both destructive and new bone formation pathogenic processes. Thus, to define the prognostic role of biomarkers in psoriasis still remains a challenge.
# Conclusions
(1) Results of the conducted studies suggest that COMP, osteoprotegerin, interleukin-20, and osteoprotegerin/ sRANKL ratio may appear useful biomarkers of bone and cartilage involvement in psoriasis.
(2) The serum concentrations of COMP, osteoprotegerin, and IL-20 were increased in the psoriatic patients.
(3) The studied biomarkers' concentrations were not significantly different in psoriatic patients without arthritis and psoriatic arthritis patients.
(4) Osteoprotegerin/sRANKL ratio was significantly lower in the psoriatic arthritis than in psoriatic patients without arthritis.
(5) Some bone and cartilage biomarkers correlated with psoriasis severity and its duration.
[fig] 1: ± 12.0 on average. The BSA values ranged from 4.0 to 85%, 27.6 ± 20.6% on average. The PGA score was 2 in 10 individuals (16.39%), 3 in 34 individuals (55.74%), 4 in another 12 individuals (19.67%), and 5 in 5 individuals (8.2%). Patients with and without arthritis did not differ significantly with respect to duration of psoriasis ( = 1.446; = 0.153), positive family history ( 2 = 0.103; = 0.747), PASI ( = 0.912; = 0.366), BSA ( = 0.703; = 0.484), and PGA ( 2 = 5.059; = 0.168). [/fig]
[table] Table 2: The biomarkers serum concentrations in the psoriatic patients compared to the control group. [/table]
[table] Table 3: The biomarkers serum concentrations in the psoriatic patients with and without arthritis. [/table]
[table] Table 4: Logistic regression analyses of the biomarkers serum concentrations in the psoriasis without arthritis and psoriatic arthritis compared to the control group. [/table]
[table] Table 5: Correlation coefficients between the biomarkers serum concentrations and clinical data in the psoriatic patients. [/table]
[table] Table 6: Mutual correlation coefficients between the biomarkers serum concentrations in the psoriatic patients. [/table]
|
Focal Lipoatrophy of Face: A Rare Esthetic Complaint
A well-proportioned face combines features that are balanced and symmetrical. Any structural alteration that leads to facial asymmetry causes esthetical and psychological disturbances. Lipoatrophy is one such condition, which results in loss of subcutaneous fat layer and manifests as a depression. Although many subtypes with varying clinical and etiological backgrounds exist, the idiopathic form is rare and facial involvement is the rarest. Computed tomography is one of the accepted diagnostic tools to determine the atrophic layer of facial anatomy. This report presents the clinical types, diagnosis, and management of a case of facial lipoatrophy.
# Introduction
Lipodystrophy broadly refers to a disturbance in the production, utilization, and storage of fat. These changes are specifically subdivided into lipoatrophy and hyperadiposity. Lipoatrophy is a sharply defined disappearance of the subcutaneous fat, without exudative reactions or appreciable fibrosis (Marble and Smith 1942). Lipoatrophy is divided into Generalized, Partial (extensive, but not generalized), and Localized (limited to a localized area) types. [bib_ref] One answer, one hundred questions, Shalev [/bib_ref] There is no universal or precise proven reason for lipoatrophy; however, literature suggests impairment of adipocyte differentiation, adipocyte apoptosis, and mitochondrial dysfunction, as the heterogeneous pathogenesis, reflecting the different subtypes. [bib_ref] Adipose tissue biology and HIVinfection, Giralt [/bib_ref] Generalized lipoatrophy is a rare disorder characterized by a near complete absence of fat. Both congenital and acquired forms are reported. The congenital type is a genetic disturbance, presenting with a generalized absence of fat within the first year of life and is followed by a series of abnormalities like insulin resistance and acanthosis nigricans before adolescence. The acquired form is autoimmune in nature. The onset is during childhood, with widespread panniculitis, and a strange physical stature.
Partial lipoatrophy was first reported in1885 by Mitchell. [bib_ref] Singular case of absence of adipose matter in the upper half of..., Mitchell [/bib_ref] It has a characteristic beginning of progressive subcutaneous fat loss in the face and scalp, and then spreads up to the iliac crests. Children between five and eight years are the target group, with girls outnumbering boys by 4:1.
Iatrogenic causes of lipoatrophy include complications of injected medications like insulin, corticosteroids, antibiotics (Penicillin G), iron, heparin, and vaccines. [bib_ref] Insulin -induced lipoatrophy in type І diabetes: A possible tumor necrosis factor-α..., Atlan-Gepner [/bib_ref] HIV-positive patients on anti-retroviral therapy, especially, nucleoside reverse transcriptase inhibitors (NRTs) [bib_ref] Adipose tissue biology and HIVinfection, Giralt [/bib_ref] and protease inhibitors (PI) [bib_ref] Subcutaneous adipocyte apoptosis in HIV -1 Protease inhibitor associated Lipodystrophy, Domingo [/bib_ref] manifest lipoatrophy as an adverse drug effect.
## Case report
A 28-year-old female patient reported to our institution with a complaint of a depression appearing on the right lower facial region. It had started as a small depression two years earlier, with no accompanying pain or discomfort, which progressed to the present state of asymmetry and abnormal appearance. Her medical records showed no history of facial trauma or dental infections. On examination, the facial asymmetry was found to be due to deficiency of the fat layer in the right parasymphyseal region, extending inferiorly to the lower border of the mandible [ [fig_ref] Figure 1a: A 28-year-old female patient with facial asymmetry and sunken right cheek [/fig_ref] ]. The regional skin was extremely thin with pigmentary changes [ [fig_ref] Figure 1b: Right profile view of the patient shows the atrophic region with pigmentation [/fig_ref] ]. The intraoral examination revealed healthy oral mucosa and a complete set of dentition. Localized idiopathic subcutaneous atrophy of the face was considered as a clinical diagnosis and the relevant diagnostic workup was started.
An en-face photograph and profile views of the patient were taken and analyzed together, to assess the degree of asymmetry [ [fig_ref] Figure 1a: A 28-year-old female patient with facial asymmetry and sunken right cheek [/fig_ref] ].
Ultrasonography using, 7-10 MHZ frequency linear probe (Volvuson 730 pro expert GE machine) revealed the normal superficial skin layer bilaterally. The thickness of subcutaneous plane was 0.22 cm on the right side and 0.28 cm on the left side. Altered echogenicity was noted on the right masseter muscle. On the affected side, the muscle thickness was 0.83 cm, whereas, on the normal side, it was 1.26 cm [ [fig_ref] Figure 1c: Ultrasound image reveals normal facial anatomy on the left side and altered... [/fig_ref].
The serial axial and coronal CT sections of the craniofacial bones obtained with multiplanar and volume-rendered reconstructions confirmed the asymmetry of the subcutaneous soft tissues, which were atrophied on the right cheek region. The underlying muscles and bones appeared normal [fig_ref] Figure 1e: Coronal CT shows subcutaneous atrophy of fat on the right side [/fig_ref].
The patient was advised biopsy, but she refused. Clinical investigations showed normal values for complete blood count, blood glucose, serum cholesterol, and triglycerides. Plasma C3 and C4 levels were within normal limit. The Antinuclear Antibody (ANA) test was negative.
# Discussion
Facial esthetics depends on the balance between the mobility and stability of facial anatomy. The subcutaneous fat provides volume and mobility, and is supported by fibrous retinacular cutis, which connects the dermis to the muscle apponeurotic system. The subcutaneous layer is critical, because, wasting of subcutaneous fat results in atrophy, whereas, its increased thickness leads to lengthening of the retinacular fibers, therefore, weakness and distention. Paucity or complete absence of fat in the confined subdermal layer, with lack of inflammatory signs, is known as idiopathic localized involutional lipoatrophy (ILIL). [bib_ref] Localized involutional lipoatrophy: Report of six cases, Yamamoto [/bib_ref] Peters and Winkelmann first reported this condition, as a sharply demarcated skin depression without any epidermal alteration. [bib_ref] Localized lipoatrophy (atrophic connective tissue disease panniculitis), Peters [/bib_ref] The features of facial lipoatrophy include sunken cheeks, deep folds in the nasolabial region, skin depression at the temples, sides of forehead, and around the eye sockets. Spontaneous regression in few Based on the clinical and imaging, we concluded the final diagnosis as localized, idiopathic involutional grade 1 lipoatrophy of the face. The patient was treated in the Plastic and Reconstructive Surgery Department. Autologous fat harvested from the abdomen was injected weeks is usual, however, in some cases the lesion may persist and cause cosmetic concerns.
The facial lipoatrophy can be graded as follows. [bib_ref] Full scope of effect of facial lipoatrophy: A framework of disease understanding, Ascher [/bib_ref] Grade 0 -No facial lipoatrophy. Grade 1 -Mild flattening or shadowing on one or more facial regions; No prominent bony land marks and no visibility of the underlying musculature. Computed tomography (CT ) plays an useful role in the diagnosis of focal lipoatrophy as fat can be easily appreciated on CT. [bib_ref] New and emerging agents in the management of lipodystrophy in HIV-infected patients, Bonnet [/bib_ref] US can demonstrate asymmetry of the face and loss of fat in focal lipoatrophy but it has less sensitivity than CT.
To improve esthetics, common treatment modalities, with evidence of safety and efficacy are, using permanent and non-permanent types of fillers. The possibility of severe infection, contour abnormalities, and facial nerve and muscle damage, restrict the use of permanent type of fillers like silicone oil. Non-permanent fillers, being biodegradable, can diminish over time and may necessitate a revamp, but they do not cause foreign body reactions or granulomas. Injection of a biodegradable and bioabsorbable poly-Llactic acid (PLA) filler causes cutaneous thickening by fibrous connective tissue formation over a month. The call for multiple treatment sessions, cost, and transitory relief, limit its use for augmentation. Autologous fat transfer (AFT) is a secure and cost-effective option, exclusively for non-HIV patients with lipoatrophy. [bib_ref] A case of fat injection for treating subcutaneous atrophy caused by local..., Imagawa [/bib_ref] In our patient, as the abnormality was a mild cosmetic disorder, with no disability, needle extraction of adipose tissue from the abdomen and injecting this at the site of the atrophy was done with the aim of ensuring rapid revascularization and less resorption. A soft, slightly curved, and well-defined anatomy with esthetic satisfaction was achieved. Frequent patient follow-up is essential to monitor the acceptance and consequences, and to see if there is a need for repeating the treatment.
# Conclusion
Facial lipoatrophy in otherwise healthy individuals is uncommon and patients usually present for cosmetic reasons. CT imaging is useful in the diagnosis and biopsy is confirmatory. This is a benign condition that can be easily corrected with surgery. The current case report highlights its diagnosis and management.
[fig] Figure 1a: A 28-year-old female patient with facial asymmetry and sunken right cheek. [/fig]
[fig] Figure 1b: Right profile view of the patient shows the atrophic region with pigmentation. [/fig]
[fig] Figure 1c: Ultrasound image reveals normal facial anatomy on the left side and altered echogenicity on the right cheek at the subcutaneous and massetric layer. [/fig]
[fig] Figure 1d: Ultrasound image of the affected side, allows measurement of muscle thickness in both open and closed mouth positions. [/fig]
[fig] Figure 1e: Coronal CT shows subcutaneous atrophy of fat on the right side. [/fig]
[fig] Figure 1f: Axial CT demonstrates subcutaneous atrophy at the level of the right mandible. [/fig]
[fig] Figure 1g: Patient's face after soft tissue reconstruction. [/fig]
[fig] Figure 1h: Patient's right profile after soft tissue reconstruction. [/fig]
[fig] Grade 2 -: Intermediate point between Grade 1 and Grade 3. Grade 3 -Moderate concavity of one or more facial regions, prominence of bony land marks, and possible visibility of the underlying musculature. Grade 4 -Intermediate point between Grade 3 and Grade 5. Grade 5 -Severe depression of one or more facial regions, severe prominence of bony land marks, and clear visibility of the underlying musculature. [/fig]
|
Dual antiplatelet therapy for reduction in mortality in patients with acute and chronic coronary syndromes
A b s t r a c tDual antiplatelet therapy (DAPT) is a foundation of successful coronary artery disease management. DAPT is recommended according to ESC guidelines for 6 months following elective percutaneous coronary intervention and for 12 months following the acute coronary syndrome (ACS). ACS are the most cost-consuming type of the ischemic heart disease, which prominently requires hospital treatment. This risk significantly increases shortly after stopping DAPT, which is typical in most patients after 12 months following acute myocardial infarction (AMI). Therefore, one of the goals of long-term treatment of such patients should be the identification of those at increased risk of subsequent events, for whom prolonged DAPT will bring clinical benefits. It has been documented that prasugrel and ticagrelor, compared to clopidogrel, significantly reduce the incidence of major adverse cardiovascular events (MACE) in ACS patients. In addition to lowering composite ischemic endpoint, ticagrelor significantly reduced all-cause and cardiovascular mortality. The long-term use of ticagrelor in patients with a previous myocardial infarction was also related to a significant reduction in MACE. In patients who had a myocardial infarction 1-3 years earlier, the addition of ticagrelor to aspirin resulted in a substantial reduction in the composite endpoint of cardiovascular death, myocardial infarction, or stroke at the expense of a small but significant increase in bleedings. Hypothetical calculations have shown that replacing clopidogrel with ticagrelor in all patients with AMI in Poland would save more than three thousand lives within 12 months after AMI and more than a thousand within the following 2 years.
## The global burden of ischemic heart disease
Ischemic heart disease is the most common cause of morbidity and mortality in Poland and worldwide. Coronary artery disease (CAD) affects around 3.5% of the population of the European Union. According to IHME data, in 2017 in Poland, 4.2% of the population suffered from CAD, which was 0.1 percentage point more than in 2012. The number of new cases of CAD in Poland decreased in the years 1990-2017. It was evident in the case of male morbidity, for whom the values in 1995 were much higher than in females, and by 2017 it had dropped to a similar level. The clinical presentation of CAD ranges from chronic (stable) coronary syndromes (CCS) to various forms of acute coronary syndromes (ACS). According to the European Society of Cardiology (ESC), two types of ACS can be distinguished: ST-segment elevation myocardial infarction (STEMI) and non-ST-segment elevation ACS (NSTE-ACS). The latter group comprises of two sub-groups: non-ST-segment elevation myocardial infarction (NSTEMI) and myocardial ischaemia without cell damage, known as unstable angina (UA). Acute coronary syndromes are the most frequent CAD presentation, which leads to death. ACS are also the most cost-consuming type of the ischemic heart disease, which prominently requires hospital treatment. In 2019, in Poland, expenses provided for ACS accounted for 61.4% of the value of hospital reimbursement due to the ischemic heart disease. In 2019, the total number of ACS cases recorded by the Polish National Health Fund was 102,700, the number was 19% lower compared to the data from 2014. This decrease mainly concerned the number of cases of unstable angina (by 55%, n = 30,000), whereas the number of myocardial infarctions increased by 9% (n = 6,000). Among women, an increase of 3% was noted, and by over 10% in men. The mean age of patients with acute coronary syndromes in 2019 was 66.3 years. Men dominated in the 60-74 age group, whereas predominance of women was noted in the group over 80 years of age. There was also an increase in the proportion of ACS patients treated with coronary angioplasty within 7 days from the onset of symptoms. Angioplasty was most often performed in the case of STEMI (88% in 2018), in a smaller percentage of NSTEMI (64%) and unstable angina (48%). A substantial number of ACS patients were treated with a coronary artery bypass grafting (CABG) procedure within 90 days of the event. The highest percentage was observed in UA patients (14%), followed by NSTEMI (6%) and STEMI patients (3%).
## Early and late outcomes following acute myocardial infarction
The nationwide AMI-PL database 2009-2012 revealed that among all AMI patients admitted in 2009, in-hospital mortality was as high as 10.5% and was strongly related to patients' age [bib_ref] Incidence, treatment, in-hospital mortality and one-year outcomes of acute myocardial infarction in..., Gierlotka [/bib_ref]. The total mortality at 1 year was 19.4%, indicating that the mortality after discharge was 8.9%. As already mentioned, AMI patients who survived the first year after the event are still at increased risk of cardiovascular events.
Wojtyniak et al. have published data from the Polish nationwide AMI-PL registry that included 134,602 patients hospitalized for acute myocardial infarction (AMI) in 2009-2010 [bib_ref] Observed and relative survival and 5-year outcomes of patients discharged after acute..., Wojtyniak [/bib_ref]. Five-year outcomes of all patients discharged alive were assessed. The AMI definition included patients with STEMI, NSTEMI and unspecified AMI. Data concerning all-cause mortality, a need for further hospitalization for cardiovascular reasons and procedures performed after discharge were collected. The outcomes were presented separately in 4 age groups: below 55, 55-64, 65-79 and above 80 years. Survival dropped the most in the first year after the index hospitalization, whereas further decline remained constant in the following years. The 5-year survival rates differ across the age groups. In males below 55 years, there was the highest survival rate, and in those above 80, the lowest one: 0.915 vs. 0.383, respectively. However, after adjustment for the overall mortality of the entire Polish population, this relationship was substantially attenuated. An even more substantial reduction in the impact of age was observed in women. However, data from the Polish registry differ from those surveyed in SWEDEHEART [bib_ref] Sex differences in treatments, relative survival, and excess mortality following acute myocardial..., Alabas [/bib_ref] and other European registries [bib_ref] Sex differences in long-term cause-specific mortality after percutaneous coronary intervention: temporal trends..., Raphael [/bib_ref] [bib_ref] Sex differences in long-term mortality after myocardial infarction: a systematic review, Bucholz [/bib_ref]. In Poland, the absolute and relative survival rate was higher in women than in men. Whereas in other countries, this correlation was the opposite. A possible explanation of this phenomenon may be caused by poor guidelines application in clinical practice, level of education, the diversity of cardiovascular risk factors and poor management of numerous comorbidities in a real-life setting. The reported high mortality rate of AMI patients after the first year following the event was consistent with other Polish data published previously. This risk significantly increases shortly after stopping dual antiplatelet therapy (DAPT), typically in most pa-tients after 12 months following AMI [bib_ref] Ischaemic risk and efficacy of ticagrelor in relation to time from P2Y12..., Bonaca [/bib_ref]. An observational study of numerous patients treated with percutaneous coronary angioplasty (PCI) for AMI showed that recurrent myocardial infarction originating from a previously untreated lesion was twice higher than from a previously stented lesion [bib_ref] 4 (66) farction: data from SWEDEHEART (Swedish Web System for Enhancement and..., Varenhorst [/bib_ref]. Therefore, one of the goals of longterm treatment of such patients should be the identification of those at increased risk of subsequent events, for whom prolonged DAPT will bring clinical benefits. At the same time, patients at high risk of bleeding should be excluded from such an approach.
It has already been demonstrated that among various risk factors, some are vital for both ischemic and bleeding events. Lindholm and coauthors studied the impact of risk factors such as age > 65, chronic kidney disease, diabetes mellitus, multivessel coronary artery disease, prior myocardial infarction and prior bleeding on the incidence of new ischemic and bleeding events [bib_ref] Combined association of key risk factors on ischaemic outcomes and bleeding in..., Lindholm [/bib_ref]. The study was based on the SWEDEHEART registry and Swedish National Patient Register that combined data obtained between 2006 and 2014. The primary outcome was the composite of MI, stroke and cardiovascular death. The majority of patients had two or more risk factors (53.5%). The authors found a strong cumulative correlation between the number of risk factors and the rate of ischemic events. With each risk factor added, there was a marked, gradual increase in the incidence of cardiovascular death, myocardial infarction or stroke. Also, a similar but less pronounced correlation with bleeding events has been found. Prior bleeding was the strongest predictor of this complication. After excluding such patients from analysis, an increasing number of risk factors remained associated with a substantial rise in the ischemic events. In contrast, the bleeding rate increased only minimally. This finding is consistent with the PRECISE-DAPT score that is recommended in ESC Guidelines for bleeding risk assessment in ACS patients and those undergoing PCI [bib_ref] ESC focused update on dual antiplatelet therapy in coronary artery disease developed..., Valgimigli [/bib_ref] [bib_ref] 2020 ESC Guidelines for the management of acute coronary syndromes in patients..., Collet [/bib_ref].
## The role of antiplatelet therapy in mortality reduction following acute myocardial infarction
Dual antiplatelet therapy (DAPT) is a foundation of successful coronary artery disease management. DAPT is recommended according to ESC guidelines for 6 months following elective PCI and for 12 months following ACS. It has been documented that more potent and predictable P2Y12 inhibitors -prasugrel and ticagrelor, compared to clopidogrel, significantly reduce the incidence of major adverse cardiovascular events (MACE) in broad-spectrum ACS patients [bib_ref] Prasugrel versus clopidogrel in patients with acute coronary syndromes, Wiviott [/bib_ref] [bib_ref] Ticagrelor versus clopidogrel in patients with acute coronary syndromes, Wallentin [/bib_ref]. In addition to lowering composite ischemic endpoint, ticagrelor significantly reduced allcause and cardiovascular mortality [bib_ref] Ticagrelor versus clopidogrel in patients with acute coronary syndromes, Wallentin [/bib_ref]. This beneficial effect has been recently confirmed in a network metaanalysis of 52,816 ACS patients from 12 randomized trials [bib_ref] Comparative efficacy and safety of oral P2Y12 inhibitors in acute coronary syndrome:..., Navarese [/bib_ref]. The long-term use of ticagrelor in patients with previous myocardial infarction was also related to a significant reduction in MACE [bib_ref] PEGASUS-TIMI 54 Steering Committee and Investigators. Long-term use of ticagrelor in patients..., Bonaca [/bib_ref]. In patients who had a myocardial infarction 1 to 3 years earlier, the addition of ticagrelor on top of a low dose of aspirin resulted in a significant reduction in the composite endpoint of cardiovascular death, myocardial infarction, or stroke at the expense of a small but significant increase in bleedings. No excess fatal bleeding or intracranial hemorrhage has been observed. In the approved European label, ticagrelor at a dose of 60 mg b.i.d. was associated with a significant reduction in cardiovascular and all-cause death -hazard ratio 0.71 (0.56-0.90) and 0.80 (0.67-0.96), respectively [bib_ref] Efficacy and safety with ticagrelor in patients with prior myocardial infarction in..., Dellborg [/bib_ref]. The EU label group consisted of patients who were randomized ≤ 2 years from qualifying MI or ≤ 1 year from stopping DAPT.
Given the well-documented positive effect of ticagrelor on mortality and its major underuse in ACS patients in Poland, it is worth calculating the hypothetical number of deaths that might be prevented by using ticagrelor instead of clopidogrel. These are theoretical calculations as it will never be possible to replace clopidogrel for everyone because of contraindications or side effects. The calculations assume the number of patients who had AMI in Poland in 2019 because the statistics for 2020, due to the COVID-19 pandemic, differ significantly from previous years. The early benefit (12 months) was calculated based on the results of the PLATO study [bib_ref] Ticagrelor versus clopidogrel in patients with acute coronary syndromes, Wallentin [/bib_ref] , whereas the long-term effect (another 24 months) was based on the approved European label, patients from PEGASUS-TIMI 54 study [bib_ref] Efficacy and safety with ticagrelor in patients with prior myocardial infarction in..., Dellborg [/bib_ref]. As shown in [fig_ref] Figure 1: A -Hypothetical reduction in mortality in patients up to 12 months after... [/fig_ref] A, the strategy of replacing clopidogrel with ticagrelor in all patients with AMI in Poland would result in more than three thousand lives saved within the 12 months following AMI. Moreover, another 1163 deaths could be avoided by adding a low dose of ticagrelor to ASA in the following 2 years [fig_ref] Figure 1: A -Hypothetical reduction in mortality in patients up to 12 months after... [/fig_ref]. These hypothetical calculations should be corrected for the actual use of antiplatelet drugs in patients with myocardial infarction in Poland. The recent analysis of the PL-ACS registry shows that in 2020, ticagrelor was used in less than 40% of patients in the first year after AMI and in less than 5% of patients in > 12 months post-AMI.
# Conclusions
In Poland, routine replacement of clopidogrel with ticagrelor would save around three thousand lives within 12 months following AMI. Adding a low dose of ticagrelor to ASA for another 24 months would prevent more than one thousand deaths.
[fig] Figure 1: A -Hypothetical reduction in mortality in patients up to 12 months after myocardial infarction. B -Hypothetical reduction in mortality in patients 12-36 months after myocardial infarction [/fig]
|
Implicit and explicit attitudes towards disease-modifying antirheumatic drugs as possible target for improving medication adherence
ObjectiveThis study aims to explore the contribution of implicit attitudes and associations towards conventional disease-modifying antirheumatic drugs (cDMARDs), alongside explicit measures, on medication-taking behaviour and clinical outcomes in adult patients with rheumatoid arthritis (RA).ResultsOf the 1659 initially-invited patients, 254 patients with RA agreed to participate in this study. Implicit attitudes correlated significantly with necessity-concerns differential (NCD) scores (ρ = 0.13, P = 0.05) and disease activity scores (ρ = -0.17, P = 0.04), whereas implicit health-related associations correlated significantly with mean scores for explicitly reported health-related associations (ρ = 0.18, P = 0.004). Significant differences in age, number of DMARDs, biologic DMARD use, NCD-scores, and self-reported correct dosing were found between the four attitudinal profiles. Nested linear regression models revealed no additional value of implicit measures in explaining self-reported medication-taking behaviour and clinical outcomes, over and above all other variables.ConclusionImplicit attitudes and associations had no additional value in explaining medication-taking behaviour and clinical outcomes over and above often used explicitly measured characteristics, attitudes and outcomes in the studied population. Only age and NCD scores contributed significantly when the dependent variable was correct dosing measured with selfreport.Implicit and explicit attitudes towards disease-modifying antirheumatic drugs PLOS ONE | https://doi.org/10.
# Methods
In this observational study, implicit attitudes (positive-negative) and health-related associations (health-sickness) were measured with Single Category Implicit Association Tests, whereas explicit outcomes were measured with a bipolar evaluative adjective scale and the Beliefs about Medicines Questionnaire Specific. The primary outcome of this study was medication-taking behaviour subjectively measured by self-report (i.e. validated Compliance Questionnaire on Rheumatology) and objectively measured with electronic drug monitors over a 3 month period. Spearman rank correlations were used to describe correlations between implicit and explicit outcomes. Nested linear regression models were used to assess the additional value of implicit measures over explicit measures and patient-, clinical-, and treatment-related characteristics.
# Introduction
Rheumatoid arthritis (RA) is a chronic inflammatory disease characterised by synovial inflammation, which can lead to irreversible articular damage, a decrease in physical functioning and quality of life, and eventually increased healthcare expenditures [bib_ref] EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological..., Smolen [/bib_ref] [bib_ref] Cost-related medication nonadherence in older rheumatoid arthritis patients, Harrold [/bib_ref] [bib_ref] What are the consequences of early rheumatoid arthritis for the individual?, Scott [/bib_ref] [bib_ref] Course of radiographic damage over 10 years in a cohort with early..., Lindqvist [/bib_ref]. Conventional diseasemodifying antirheumatic drugs (cDMARDs) are the cornerstone of RA treatment and are fundamental to prevent radiologic progression on long-term [bib_ref] EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological..., Smolen [/bib_ref]. Maximum treatment benefits can, however, only be achieved if patients adhere to their treatment [bib_ref] Electronic monitoring of oral therapies in ethnically diverse and economically disadvantaged patients..., Waimann [/bib_ref]. Previous research in patients with RA revealed that adherence rates to DMARDs varied from 30% to 107%, depending on the used measurement method [bib_ref] Adherence to disease modifying anti-rheumatic drugs in rheumatoid arthritis patients: a narrative..., Salt [/bib_ref]. So far, interventions designed to improve medication adherence were only partly effective in changing medication-taking behaviour [bib_ref] What are the effects of medication adherence interventions in rheumatic diseases: a..., Galo [/bib_ref] [bib_ref] Identification and assessment of adherence-enhancing interventions in studies assessing medication adherence through..., Demonceau [/bib_ref] [bib_ref] Methods to improve medication adherence in patients with chronic inflammatory rheumatic diseases:..., Lavielle [/bib_ref]. An explanation for the ineffectiveness of adherence-improving interventions might be that previous studies have largely focused on patient's explicit, 'conscious', evaluations of e.g. medication or medication-taking behaviour [bib_ref] Implicit processes, self-regulation, and interventions for behavior change, St Quinton [/bib_ref]. These interventions are often designed on the basis of theories such as the theory of planned behaviour and the health belief model, which form the backbone for understanding how explicit evaluations affect behaviour [bib_ref] Implicit processes, self-regulation, and interventions for behavior change, St Quinton [/bib_ref] [bib_ref] Historical Origins of the Health Belief Model, Rosenstock [/bib_ref]. However, extensive theoretical and empirical contributions in the field of psychology demonstrate that behaviour is only partly driven by conscious, explicit evaluations [bib_ref] Implicit processes, self-regulation, and interventions for behavior change, St Quinton [/bib_ref] [bib_ref] Implicit attitudes and their measurement: theoretical foundations and use in consumer behavior..., Ackermann [/bib_ref] [bib_ref] Nonconscious processess and health, Sheeran [/bib_ref] [bib_ref] Nature and operation of attitudes, Ajzen [/bib_ref]. A lot of behaviour originates from more subconscious or automatic processes (i.e. implicit associations) [bib_ref] Implicit processes, self-regulation, and interventions for behavior change, St Quinton [/bib_ref] [bib_ref] Implicit attitudes and their measurement: theoretical foundations and use in consumer behavior..., Ackermann [/bib_ref] [bib_ref] Nonconscious processess and health, Sheeran [/bib_ref] [bib_ref] Nature and operation of attitudes, Ajzen [/bib_ref]. Dual process theories, which account for both reflective (conscious) as well as automatic (subconscious) drivers of behaviour, are well-accepted in various scientific domains aimed at unravelling the mechanisms underlying human behaviour (i.e., psychology, behavioural economics, marketing, communication). Adherence research, in contrast, has rarely tapped into this knowledge base, generally ignoring patients' automatic associations with their medication or medication-taking behaviour. The way to tap into these automatic processes is to infer people's associations from speeded response tasks (implicit measurement), rather than ask them to introspect (explicit measurement) [bib_ref] Attitudes and Attitude Change, Bohner [/bib_ref] [bib_ref] Measuring individual differences in implicit cognition: the Implicit Association Test, Greenwald [/bib_ref]. In this study, we used such implicit measurements with the aim to tap into automatically activated associations that are based on a past experience and mediate favourable or unfavourable feelings that individuals may not be aware of. We refer to these associations as implicit attitudes, following the evidence that implicit measurements are capable of exposing automatic positive or negative associations, but acknowledging that they do not by definition do so and that implicit refers to the measurement rather than to the process per se [bib_ref] Attitudes and Attitude Change, Bohner [/bib_ref] [bib_ref] Implicit measures: a normative analysis and review, Houwer [/bib_ref]. Explicit attitudes, on the contrary, are defined as deliberate or (sub)conscious evaluations of medication.
To date, few studies have investigated implicit attitudes towards medication and their potential relation to medication-taking behaviour [bib_ref] Disentangling rheumatoid arthritis patients ' implicit and explicit attitudes toward methotrexate, Linn [/bib_ref] [bib_ref] Implicit versus explicit attitudes toward psychiatric medication: implications for insight and treatment..., Rüsch [/bib_ref]. No strong correlation between implicit and explicit attitudes towards medication was previously found, nor between implicit attitudes and self-reported (thus explicitly measured) medication-taking behaviour [bib_ref] Disentangling rheumatoid arthritis patients ' implicit and explicit attitudes toward methotrexate, Linn [/bib_ref] [bib_ref] Implicit versus explicit attitudes toward psychiatric medication: implications for insight and treatment..., Rüsch [/bib_ref]. This was no surprise, given that 1) implicit and explicit processes are known to deviate (cf. dual process models and findings), and 2) self-report to assess medication-taking behaviour is susceptible to the limits of patient's self-knowledge and the tendency to provide socially desirable answers, and may therefore deviate from actual behaviour [bib_ref] Adherence to Medication, Osterberg [/bib_ref]. Assessing medicationtaking behaviour based on self-reports is therefore an important limitation in previous research on implicit and explicit attitudes [bib_ref] Disentangling rheumatoid arthritis patients ' implicit and explicit attitudes toward methotrexate, Linn [/bib_ref] [bib_ref] Implicit versus explicit attitudes toward psychiatric medication: implications for insight and treatment..., Rüsch [/bib_ref]. It can be argued that explicit determinants assessed with self-report correlate more strongly with self-reported behaviour (because both are assumed to be driven by elaborative thought), whereas implicit attitudes may have the potential to correlate uniquely with actual behaviour, which is known to be driven largely by automatic or subconscious processes [bib_ref] Attitudes and Attitude Change, Bohner [/bib_ref] [bib_ref] Disentangling rheumatoid arthritis patients ' implicit and explicit attitudes toward methotrexate, Linn [/bib_ref]. Objective measurement methods, such as electronic drug monitors, are therefore more suitable to examine implicit attitudes' relation with actual medication-taking behaviour.
Therefore, the primary objective of this study is to explore the contribution of implicitly measured attitudes towards cDMARDs, alongside explicit measures, on medication-taking behaviour (i.e. measured subjectively with self-report and objectively with electronic drug monitors) of patients with RA. The secondary objective is to explore the contribution of implicitly measured attitudes towards cDMARDs, alongside explicit measures, on disease activity scores in patients with RA.
# Methods
## Study design and setting
An observational study was conducted in two of the largest specialised rheumatology centres across the Netherlands (i.e. covering approximately 20% of all patients with RA): Sint Maartenskliniek (Nijmegen) and Reade (Amsterdam). Participants were recruited between July 5 th , 2016 and November 30 th , 2017. The STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement for observational studies and EMERGE (ESPACOMP Medication Adherence Reporting Guideline) were used as guidance for adequate reporting of this study (See S1 Table for all the abbreviations used in this study) [bib_ref] The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines..., Elm E Von Altman [/bib_ref] [bib_ref] ESPACOMP Medication Adherence Reporting Guideline (EMERGE), Geest De [/bib_ref].
## Eligibility criteria and patient selection
Consecutive adult (� 18 years) patients with a clinical diagnosis of RA, treated with at least one cDMARD for a minimum period of one year were invited to participate in this study. No additional in-and exclusion criteria were defined. Four weeks before their planned regular consultation with their treating clinician, patients received written information and an informed consent form. After one to two weeks, patients were approached by telephone to ask about their willingness to participate in this study. In case patients agreed to participate, the researcher planned a research appointment before their regular consultation in order to sign the informed consent form and complete baseline measurements.
## Study outcomes
Primary outcome of this study was medication-taking behaviour measured over a period of three months after inclusion. Both self-report (i.e. the validated Compliance Questionnaire on Rheumatology (CQR)) and electronic drug monitors (i.e. Medication wAardex1)) were used to assess the implementation of the dosing regimen [bib_ref] The Compliance-Questionnaire-Rheumatology compared with electronic medication event monitoring: a validation study, De Klerk [/bib_ref] [bib_ref] Development of a questionnaire to investigate patient compliance with antirheumatic drug therapy, De Klerk [/bib_ref] [bib_ref] A scoping review of studies comparing the medication event monitoring system (MEMS)..., Alili [/bib_ref]. The operational definition for the implementation of dosing regimens was correct dosing, which is defined as the proportion of days with the correct number of doses taken [bib_ref] The Compliance-Questionnaire-Rheumatology compared with electronic medication event monitoring: a validation study, De Klerk [/bib_ref] [bib_ref] A new taxonomy for describing and defining adherence to medications, Vrijens [/bib_ref]. Patients were considered to be adherent if the proportion of correct days was >80% based on the prescribed medication regimen by the health professional. Continuous adherence data were used for nested linear regression models. Patient's disease activity score (DAS28-CRP) was the secondary outcome measure, which was assessed in conformity with treatment protocols as part of standard care [bib_ref] EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological..., Smolen [/bib_ref].
## Variables and data collection
Baseline measurements were performed in the consulting room of the outpatient pharmacy. The following data were collected with a hardcopy questionnaire at baseline: socio-demographic characteristics (i.e. age, sex, educational level, living status, ethnicity), explicit attitudes and health-related associations with cDMARDs with a bipolar evaluative adjective scale (8 items to evaluate positive-negative attitudes, 10 items to evaluate health-sickness related associations), beliefs about medicines with the Beliefs about Medicines Questionnaire Specific (BMQ-Specific, 10 Likert-scaled items, ranging from 1 to 5), and self-reported medication-taking behaviour with the validated CQR with 19 Likert-scaled (ranging from 1 to 4) items [bib_ref] The Compliance-Questionnaire-Rheumatology compared with electronic medication event monitoring: a validation study, De Klerk [/bib_ref] [bib_ref] Development of a questionnaire to investigate patient compliance with antirheumatic drug therapy, De Klerk [/bib_ref] [bib_ref] The beliefs about medicines questionnaire: the development and evaluation of a new..., Horne [/bib_ref]. Implicit attitudes and health-related associations were collected with Single Category Implicit Association Tests (SC-IATs), which were performed on a laptop in the consulting room of the outpatient pharmacy [bib_ref] Measuring individual differences in implicit cognition: the Implicit Association Test, Greenwald [/bib_ref]. Clinical characteristics (i.e. disease duration, serology, type and current number of DMARD(s), disease activity scores) were extracted from patient's medical file by the local researchers. Patient diaries were handed at baseline in order to let patients register possible unintended openings of MEMS. Follow-up measurements were performed at the day of the next planned regular consultation, with a minimum and maximum interval between baseline-and follow-up visits of three and nine months, respectively. At follow-up visit, MEMS read-outs were used to assess correct dosing over the previous months.
## Measurement instruments
Single category implicit association tests (SC-IATs). SC-IATs (Inquisit version 5) were used to measure automatic associations [bib_ref] Measuring individual differences in implicit cognition: the Implicit Association Test, Greenwald [/bib_ref] [bib_ref] Understanding and using the Implicit Association Test: III. Meta-Analysis of predictive validity, Greenwald [/bib_ref]. The SC-IAT is considered a reliable and valid instrument to measure implicit associations, which in this study is constituted of two concepts, i.e. attitudes towards cDMARDs (positive versus negative) and health-related associations with cDMARDs (health versus sickness). The measurement of each concept included three rounds: one practice round of 20 trials followed by two experimental rounds of 40 trials each (see S1 File). Each trial was defined as a computerised categorisation task in which automatic associations were measured based on patient's response times. The response times in the experimental rounds serve as a proxy for association strength, where faster responses represent stronger associations. For instance, if patients were on average faster in trials coupling drug stimuli and positive (versus negative) stimuli, this reflects a relatively positive (versus negative) automatic association with cDMARDs. S1 File provides a more detailed description of the SC-IATs procedures used in this study. The design and procedures of the SC-IATs were based on the pilot study of Linn et al [bib_ref] Disentangling rheumatoid arthritis patients ' implicit and explicit attitudes toward methotrexate, Linn [/bib_ref].
Explicit medication attitudes, health-related associations and beliefs about medicines. Explicit attitudes (10 items, e.g. I think [name cDMARD] is 1 negative-5 positive) and explicit health-related associations (8 items, e.g. 'to what extent do you associate [name cDMARD] with the following terms', 1 dead-5 alive) were assessed with a bipolar evaluative adjective scale. The items were identical to those in the SC-IATs (see S2 File for the complete list of words used in the bipolar evaluative adjective scale). Medication necessity and concern beliefs were assessed with the validated Beliefs about Medicines Questionnaire (BMQ) Specific (10 items, 5 necessity items and 5 concern items) [bib_ref] The beliefs about medicines questionnaire: the development and evaluation of a new..., Horne [/bib_ref]. Item scores varied from 1 (strongly disagree) to 5 (strongly agree), which resulted in sum scale scores of 5 to 25 for each subscale (necessity beliefs versus concern beliefs).
## Study size
A common rule of thumb is to formulate sample size requirements as events per variable, with a minimum of 10 events per variable. Assuming a sample size requirement of 10 non-adherent patients per variable and a prevalence of 33% of non-adherence, a sample of 240 patients is sufficient to build a reliable logistic model including a maximum of 8 independent variables. Taking into account 15% loss to follow-up, a sample size of 275 patients was required.
# Statistical methods
Statistical analyses were performed with STATA version 13.1. Descriptive statistics were used to describe patient characteristics by using mean (SD) or median (P25-P75) depending on the distribution of measurements. Educational level was classified in low, moderate or high educational level. Low educational level was defined as no education, (extended) primary education or pre-vocational education, moderate educational level was defined as vocational education or selective secondary education, and high educational level was defined as education provided by universities of applied sciences and research universities. Data were presented as percentages in case of proportions. P-values �0.05 were considered statistically significant.
Implicit attitudes and health-related associations were assessed in terms of response times in milliseconds (ms) on the SC-IATs. The improved IAT scoring algorithm described by Greenwald et al. was used as scoring procedure to calculate the D measure for strength of automatic associations. S1 File provides a detailed description of the statistical methods to calculate D measures. D measures above 0 indicated that patients had relatively faster response times on positive categorisation rounds than on negative categorisation rounds, and were interpreted as a relatively more positive than negative implicit attitude towards cDMARDs or a relatively more health-related association than sickness-related association, and vice versa.
For explicit attitudes, mean sum scale scores were calculated for each concept (i.e. attitudes and health-related associations). Patients with a mean scale score below or equal to 3 were considered to be relatively negative or have sickness-related associations with cDMARDs, whereas patients with mean scale scores higher than 3 were considered to be relatively positive or associated their cDMARD use with health. Beliefs about medicines were operationalised as necessity-concerns differential (NCD) scores [bib_ref] The beliefs about medicines questionnaire: the development and evaluation of a new..., Horne [/bib_ref] [bib_ref] Perceived need to take medication is associated with medication non-adherence in patients..., Zwikker [/bib_ref]. This NCD was calculated by subtracting the sum of the item scores for concerns from the sum of item scores for necessity beliefs. A negative NCD indicated that concern beliefs predominate necessity beliefs and vice versa [bib_ref] The beliefs about medicines questionnaire: the development and evaluation of a new..., Horne [/bib_ref] [bib_ref] Perceived need to take medication is associated with medication non-adherence in patients..., Zwikker [/bib_ref].
Participants were also categorised in attitudinal profiles based on their (in)congruent (implicit and explicit) attitudes and health-related associations based on the cut-off scores for implicit and explicit attitudes and health-related associations as described above. Depending on the distribution and type of variables, Two-sample t-tests, Pearson chi-square tests, Fisher's exact tests, Bonferroni corrected post-hoc tests, and proportion tests were performed to test for significant differences in patient characteristics between study sites and attitudinal profiles.
Self-reported medication-taking behaviour was calculated with the discriminant function for CQR items as described by de Klerk et al [bib_ref] The Compliance-Questionnaire-Rheumatology compared with electronic medication event monitoring: a validation study, De Klerk [/bib_ref]. The critical cut-off score of -2.0046 for correct dosing �80% was used to identify adherent and non-adherent patients [bib_ref] The Compliance-Questionnaire-Rheumatology compared with electronic medication event monitoring: a validation study, De Klerk [/bib_ref]. Medicationtaking behaviour measured with self-report was compared with medication-taking behaviour measured with MEMS. The percentage of adherent patients per day was calculated over time during three months follow-up and presented per attitudinal profile. The proportion of patients in remission was presented for each attitudinal profile based on the cut-off scores for DAS28-CRP and DAS28-BSE, as described by Fleischmann et al.
Spearman rank correlations were used to describe the correlation between patient's implicit and explicit outcomes, medication-taking behaviour (i.e. measured by self-report and using MEMS), and clinical outcomes. Because of the explorative (rather than hypothesis-testing) character of this study, no multiple testing corrections were performed over the separate correlational analyses. Nested linear regression models (i.e. by sequentially adding blocks of variables) were used to assess the additional value of implicit measures, over explicit measures and patient characteristics, in explaining adherence to medication and disease activity scores in patients with RA. For each separate model three dependent continuous variables were used: correct dosing based on (1) the CQR discriminant function, and (2) MEMS data, followed by (3) DAS28-CRP scores at baseline.
# Ethical approval
This study was conducted according to the ethical principles for medical research as stated in the Declaration of Helsinki (64 th WMA General Assembly, Fortaleza, Brazil, October 2013) and was approved by the Medical Research Ethics Committee of Arnhem-Nijmegen (File: 2016-2410).
## Patient and public involvement
Two patient research partners were involved in the design phase of this study. Those patient research partners pretested the Single Category Implicit Association Tests and assessed the comprehensibility of the hardcopy questionnaire for patients with rheumatoid arthritis.
# Results
## Study sample characteristics
Of the 1,659 initially invited patients, 254 patients agreed to participate in this study. The overall response rate was 15.3% (Nijmegen: 15.4%; Amsterdam: 15.0%). [fig_ref] Fig 1: Flowchart of study participants [/fig_ref] an overview of patient recruitment, patient inclusion and drop-outs during follow-up. Participants had a mean age of 62.8 (SD:11.2) years, 68.1% was female, 32.9% of the patients was highly educated, and 22.2% was living alone. The mean disease duration of patients was 11.8 (SD:9.0) years and biologic DMARDs were prescribed to 32.7% of the included patients. Methotrexate tablets were significantly more often prescribed in Amsterdam compared to Nijmegen (resp. 77.5% versus 40.4%, P<0.0001), whereas methotrexate subcutaneous injections were significantly more often prescribed in Nijmegen than in Amsterdam (resp. 31.1% versus 7.0%, P = 0.0001). [fig_ref] Table 1: Study sample characteristics [/fig_ref] provides an overview of all patient characteristics. Due to different treatment protocols applied at both study sites, DAS28-CRP scores were only available for patients treated in Nijmegen (N = 152).
Of the 254 patients who agreed to participate, 98.8% completed all SC-IATs, 98.0% the bipolar evaluative adjective scale, 98.8% the BMQ-Specific, 99.2% the CQR and 91.7% of the patients provided MEMS data. The main reason for missing MEMS data was discontinuation of cDMARD therapy [fig_ref] Fig 1: Flowchart of study participants [/fig_ref]. Continuous CQR adherence data instead of dichotomous data were used in this study, since the validation of the CQR against MEMS together with the arbitrary cut-off score of 80% has been subject to considerable debate [bib_ref] Nonadherence to disease modifying antirheumatic drugs in the first year after diagnosis:..., Pasma [/bib_ref] [bib_ref] Determinants of methotrexate adherence in rheumatoid arthritis patients, De Cuyper [/bib_ref]. The proportion of correct trials in the SC-IAT experimental rounds for the positive-negative concept was 93.3% and for the health-sickness concept 95.5%. No participants were excluded for further dataanalysis of implicit data.
## Patient's implicit and explicit attitudes and associations, including beliefs about medication
The mean D measure for implicit attitudes towards cDMARDs was -0.054 (SD: 0.42, range: -1.26; 1.21, skewness: -0.11, kurtosis: 2.81), whereas the mean D measure for implicit healthrelated associations with cDMARDs was -0.10 (SD: 0.38, range: -1.25; 1.03, skewness: -0.03, kurtosis: 3.24). The mean score for explicit attitudes of patients who completed the bipolar evaluative adjective scale was 3.5 (SD:0.72, range: 1.4; 5, skewness: -0.24, kurtosis: 3.25), which was similar for explicit health-sickness associations with cDMARDs (M = 3.6, SD: 0.89, range: 1; 5, skewness: -0.26, kurtosis: 2.47). Regarding beliefs about medicines, the mean sum scale score for necessity beliefs and for concern beliefs was 19.9 (SD: 3.6, range: 5; 25, skewness: -0.78, kurtosis: 3.9) and 14.1 (SD: 3.9, range: 5; 25, skewness: -0.18, kurtosis: 2.58), respectively. The mean necessity-concerns differential score of 5.8 (SD: 5.2, range: -16; 19, skewness: -0.17, kurtosis: 3.8) indicates that necessity beliefs outweigh patient's concerns about medication.
Patient's implicit attitudes (D measure) and explicit attitudes (mean score measured with the bipolar evaluative adjective scale) were not significantly correlated (ρ = 0.09, P = 0.16), however a significant weak correlation was found between mean D measures for implicit attitudes and necessity-concerns differential scores (ρ = 0.13, P = 0.05, illustrated in [fig_ref] Fig 2: Fig 2 [/fig_ref]. Also implicit health-related associations were significantly (yet weakly) correlated with mean scores for explicitly reported health-related associations (ρ = 0.18, P = 0.004). No significant Explicit attitudes (measured with a bipolar evaluative adjective scale) were not correlated with self-reported correct dosing (ρ = 0.10, P = 0.12), MEMS correct dosing (ρ = 0.10, P = 0.13), and disease activity scores (ρ = -0.09, P = 0.30). Also explicit health-related associations were not correlated with MEMS correct dosing (ρ = 0.11, P = 0.08), however, a significant negative correlation was found between explicit health-related associations and disease-activity scores (ρ = -0.21, P = 0.01), and a significant positive correlation with self-reported correct dosing (ρ = 0.13, P = 0.05). NCD-scores were significantly correlated with self-reported correct dosing (ρ = 0.26, P<0.0001), but not with MEMS correct dosing and disease activity scores.
## Congruent and incongruent (implicit and explicit) attitudinal profiles
Fewer patients displayed implicit positive attitudes (47.2%) and health-related associations (39.6%) regarding cDMARD therapy than when explicitly asked (75.6% and 67.3%, respectively). Based on D measures for implicit attitudes and mean scores for explicit attitudes from the bipolar evaluative adjective scale, patients were categorized in four attitudinal profiles per concept, indicating the (in)congruently positive or negative nature of their implicit and explicit attitudes (see [fig_ref] Table 2: Description of profiles of patients diagnosed with rheumatoid arthritis based on [/fig_ref] for an overview of subgroup characteristics for each attitudinal profile for the concept positive-negative). About half of all participants displayed incongruent (implicit and explicit) attitudes and health-related associations (46.3% and 49.0%, respectively). Bonferroni corrected post-hoc tests revealed that congruently negative patients (profile III) differed significantly in age, NCD-scores, bDMARD use and self-reported correct dosing from other attitudinal profiles. Congruently negative patients were on average 6.1 years younger (95%CI: -11.74; -0.46, P = 0.03) than congruently positive patients, reported more concerns than explicitly positive patients (contrast with profile I:
## Correlation between implicit attitudes (d measure) and explicit beliefs about medication (necessityconcerns differential scores).
Higher NCD-scores indicate that necessity beliefs outweigh concern beliefs, whereas higher D measures for implicit attitudes indicate more positive attitudes towards cDMARDs (ρ = 0.13, P = 0.05). Abbreviations: cDMARD (conventional disease-modifying antirheumatic drug), NCD (necessity-concerns differential). time between different profiles, similar patterns emerge within the four attitudinal profiles [fig_ref] Fig 3: The proportion of adherent patients with rheumatoid arthritis over time based on... [/fig_ref]. However, the proportion of adherent patients with negative explicit attitudes (profiles III and IV) more often dropped below the 80% level than for the positive explicit attitudinal profiles (I and II). The small number of patients categorised in these explicitly negative profiles might have contributed to these findings. Additionally, the largest difference in proportion of adherent patients between both measurement instruments (self-report versus MEMS) was found for patients assigned to profile IV. No significant differences between attitudinal profiles were found in the number of comorbidities and the total number of DMARDs after Bonferroni adjusted post-hoc tests. Subgroup characteristics for each profile regarding health-related associations can be found in S2 Patients who displayed congruently sickness-related associations with cDMARDs differed significantly in NCD-scores from patients who reported explicit (regardless of their implicit) health-related associations (contrast with profile I: -4.1, P<0.001; contrast with profile II: -3.0, P = 0.002). In other words, patients who displayed congruent (implicit and explicit) sickness-related associations with cDMARDs had more concerns about their medication than patients who reported explicit health-related associations with cDMARDs. The same applied for patients categorised in profile IV compared with patients categorised in profile I (contrast: -6.2, P<0.001) and profile II (contrast:-5.0, P<0.001). The mean number of DMARDs in the congruently sickness-related subgroup was significantly higher than the mean number of DMARDs in the congruently health-related subgroup (contrast: 0.29, P = 0.04). When exploring the proportion of adherent patients measured with MEMS over time between the different health-related attitudinal profiles, results were similar to those of the positive-negative concept (data not shown).
## Added value of implicit attitudes in explaining medication-taking behaviour and clinical outcomes
The skewed MEMS data (i.e. high proportion of adherent patients), even after data transformation, did not meet the assumptions for linear regression modelling. Medication-taking behaviour measured with MEMS was, therefore, excluded for nested linear regression modelling. Nested linear regression models showed that only age and NCD-scores contributed significantly when the dependent variable was correct dosing measured with self-report. However, no variables measured in this study contributed significantly to disease activity scores. Implicit attitudes and implicit health-related associations, alongside patient characteristics and explicit measures, had no added value in explaining variance in medication-taking behaviour when measured with self-report. The same applied for disease activity scores (see [fig_ref] Table 3: Results nested linear regression models [/fig_ref] and [fig_ref] Table 4: Results nested linear regression models [/fig_ref]. Overall, determinants measured in this study explained approximately 11 to 17% of the variance in medication-taking behaviour and disease activity scores.
# Discussion
This study showed that implicit attitudes and health-related associations do not explain medication-taking behaviour and clinical outcomes in clinical practice over and above 1) patient characteristics, 2) clinical variables, 3) treatment characteristics, 4) explicit attitudes and health-related associations, or 5) beliefs about medicines. Only age and NCD-scores contributed significantly to medication taking behaviour measured with self-report. However, some significant but weak associations were found between implicit attitudes and necessity-concerns differential scores, between implicit and explicit health-related associations, and implicit attitudes and disease activity. The categorization of patients in combined implicit and explicit attitudinal profiles revealed that characteristics of in particular patients with congruently negative or congruently sickness-related associations, differed from other attitudinal profiles.
To our knowledge, this is the first study which compares groups of patients with congruent and incongruent implicit and explicit attitudes towards medication in relation to objectively measured medication-taking behaviour and clinical outcomes in patients with RA. Linn et al already emphasized the importance of exploring implicit associations as possible targets for improving actual (rather than self-reported) medication adherence in this population [bib_ref] Disentangling rheumatoid arthritis patients ' implicit and explicit attitudes toward methotrexate, Linn [/bib_ref]. In terms of (in)congruent implicit and explicit attitudes towards medication, our findings are in line with the findings reported by Linn et al and Rüsch et al, which supports the idea that implicit and explicit measures are different but related constructs [bib_ref] Disentangling rheumatoid arthritis patients ' implicit and explicit attitudes toward methotrexate, Linn [/bib_ref] [bib_ref] Implicit versus explicit attitudes toward psychiatric medication: implications for insight and treatment..., Rüsch [/bib_ref]. However, contrary to our expectations, patient's implicit associations did not significantly explain variance in objectively measured medication-taking behaviour. Several possible explanations can be given to elucidate this finding.
One explanation is that no association between implicit associations and objectively measured medication-taking behaviour in clinical practice exist. In other words, implicit associations (with cDMARDs) do not underlie (medication-taking) behaviour. This finding is contrary to previous studies [bib_ref] Implicit processes, self-regulation, and interventions for behavior change, St Quinton [/bib_ref] [bib_ref] Predictive validity of the implicit association test in studies of brands, consumer..., Maison [/bib_ref] [bib_ref] Implicit measures in social cognition research: their meaning and use, Fazio [/bib_ref]. However, the concept of predicting behaviour by implicit measures was recently challenged by studies demonstrating that correlations between implicit measures and measures of behaviour are often small to medium [bib_ref] What do implicit measures measure, Brownstein [/bib_ref] [bib_ref] Attitudes and Behavior, Guyer [/bib_ref] [bib_ref] Predicting ethnic and racial discrimination: a meta-analysis of IAT criterion studies, Oswald [/bib_ref]. Person-, context-, and behaviour-specific moderators might be responsible for these contradictory findings [bib_ref] What do implicit measures measure, Brownstein [/bib_ref] [bib_ref] Attitudes and Behavior, Guyer [/bib_ref]. For instance, medication-taking behaviour can be seen as a more habitual, recurrent behaviour that might initially originate in conscious thought processes ("I should not forget my medication"), whereas some of the other examined behaviours can be argued to be less habitual and frequent (e.g., voting) and less rooted in conscious thought (e.g., brand preferences). The second explanation for the current findings might be that implicit processes do underlie objectively measured medication-taking behaviour, but were not detected in our study due to methodological limitations, described below.
# Methodological considerations
The key strengths of this study are the large sample size and patient recruitment in two of the largest rheumatology specialized centres across the Netherlands (i.e. covering approximately 20% of all patients with RA). Another strength is the use of MEMS to objectively measure medication-taking behaviour over three months in addition to the assessment of medicationtaking behaviour by self-report. Measuring medication-taking behaviour by self-report is more susceptible to recall bias [bib_ref] Adherence to Medication, Osterberg [/bib_ref]. In contrast with Rüsch et al, we used validated questionnaires to assess self-reported adherence and beliefs about medicines [bib_ref] Implicit versus explicit attitudes toward psychiatric medication: implications for insight and treatment..., Rüsch [/bib_ref]. An advantage of using both (validated) questionnaires and MEMS is the ability to compare both measurement instruments. However, MEMS is considered as gold standard method, since this method Implicit and explicit attitudes towards disease-modifying antirheumatic drugs provides insight in more objectively measured daily medication-taking behaviour over time compared with self-report at one time point. Still, it is an assumption that MEMS device usage corresponds with actual medication intake since intentional non-adherence to medication (e.g. opening MEMS without taking the medication) cannot be prevented in the home setting of patients. The awareness of being monitored (i.e. Hawthorne effect), might have contributed to the large proportion of adherent patients (i.e. the small amount of variation in adherence measures, in particular when MEMS are used) that we have found in our study when medication-taking behaviour was measured with both self-report and electronic drug monitors. This small amount of variation might have limited the power of this study to detect differences between adherent and non-adherent patients across attitudinal profiles, since high adherence rates could indicate more conscious and rational (e.g. planned) intake behaviour or high patient engagement with their treatment, in which case explicit attitudes and beliefs about medicines would override (potentially different) implicit associations. As a consequence, this might also limit the generalizability of the results. The validity of the SC-IATs used in this study might be questioned since the target population might have a limited hand function, which might provide insufficient contrast between the experimental rounds in the SC-IATs. Also, it is unclear if the words and pictures used as stimuli in the SC-IATs are optimally related to patient's medication use [bib_ref] Attitudes and Behavior, Guyer [/bib_ref]. However, pictures were created based on pharmacy records in participating centres (i.e. manufacturer of the drugs, type of packaging, and appearance of the drug) and SC-IATs were personalised based on patient's cDMARD use to increase the ability of patients to recognise their cDMARD at a glance. Another methodological consideration was the cut-off score for categorizing patients in attitudinal profiles based on (scale) midpoints for positive-negative attitudes and health-sickness related associations. The large proportion of patients who displayed relatively neutral implicit attitudes and associations (i.e. D measure near midpoint 0) or neutral explicit attitudes or health-related associations (i.e. mean score near midpoint 3) might have reduced the contrast in characteristics (e.g. patient-, clinical-,and treatment-related variables, beliefs about medicines, medication-taking behaviour, and clinical outcomes) across attitudinal profiles. Currently, there is discussion in the literature on whether implicit measures by definition expose automatic associations or subconscious attitudes or whether they might also be sensitive to more reflective and conscious thought [bib_ref] Implicit measures: a normative analysis and review, Houwer [/bib_ref] [bib_ref] What do implicit measures measure, Brownstein [/bib_ref]. There is no consensus yet on the terminology and nature of these implicit and explicit processes and measurements, on how these constructs interact with each other, and how stable these constructs are over time and across situations [bib_ref] Attitudes and Attitude Change, Bohner [/bib_ref] [bib_ref] Implicit measures: a normative analysis and review, Houwer [/bib_ref] [bib_ref] What do implicit measures measure, Brownstein [/bib_ref]. Future research should take these considerations into account. Implicit and explicit attitudes towards disease-modifying antirheumatic drugs
## Generalizability of the results
The external validity of the results might be questionable due to the low response rate of patients, together with the high proportion of adherent participants. The latter might indicate selection bias since it is well known that in adherence research often only highly motivated patients are willing to participate in studies [bib_ref] Adherence intervention research: what have we learned and what do we do..., Bender [/bib_ref]. It is conceivable that our findings on MEMS adherence are, therefore, an overrepresentation of adherence rates in the general RA population. Also the large proportion of patients who had a Dutch ethnic background in a study site which is located in a multicultural setting, together with the long disease duration, support the idea of selection bias. Since we did not measure health literacy in this study, it is unknown if patients with low health literacy skills were underrepresented in our study sample. However, when comparing study participants to the general RA population in the Sint Maartenskliniek (chosen as reference due to limited access of data of non-participating patients across study sites), no significant differences were found in mean age, sex, and mean disease duration. The proportion of study participants using a biologic DMARD was smaller than the proportion of bDMARD users in the general population in the Sint Maartenskliniek (32.7% and 41.5% respectively), whereas the proportion of patients who are in remission was higher in the general RA population in the Sint Maartenskliniek compared to our study sample (71.5% and 41.7% respectively). The latter might also explain the large proportion of adherent patients, since patients with high disease activity scores might be more motivated to participate in this study and might also be more likely to adhere to their treatment during the study. It is also assumed that our findings on patient's implicit and explicit attitudes towards cDMARDs cannot simply be extrapolated to biologic DMARDs and the recently introduced JAK-inhibitors. Taking together, our study population might not be optimally representative for the entire RA population, especially for patients with early RA, and ethnic minorities.
In conclusion, about half of the patients with RA showed incongruent (implicit and explicit) attitudes and health-related associations with cDMARDs. Implicit attitudes and associations had no additional value in explaining medication-taking behaviour and clinical outcomes over and above often used explicitly measured characteristics, attitudes and outcomes in the studied population. However, this research provides interesting areas for future research regarding implicit and explicit processes that might be involved in medication-taking behaviour. Attitudinal profiles: Health versus sickness associations with cDMARDs. Description of profiles of patients diagnosed with rheumatoid arthritis based on (in)congruent implicit and explicit health-versus sickness-related associations with cDMARDs. Categories were low versus medium-high for educational level, living alone versus living together (with children and/or partner) for residential status, and for ethnic background Dutch versus other. Variables with P-values �0.05 (unadjusted for multiple testing) were further analyzed with Bonferroni corrected post-hoc tests. Abbreviations: DAS28-CRP (Disease Activity Score based on 28 joints and C-Reactive Protein), DMARD (disease-modifying antirheumatic drug), bDMARD (biologic DMARD), MEMS (Medication Event Monitoring System). (PDF)
## Supporting information
[fig] Fig 1: Flowchart of study participants. https://doi.org/10.1371/journal.pone.0221290.g001 Implicit and explicit attitudes towards disease-modifying antirheumatic drugs PLOS ONE | https://doi.org/10.1371/journal.pone.0221290 August 30, 2019correlation was found between implicit health-related associations and NCD-scores (ρ = 0.09, P = 0.18).Implicit attitudes (D measure) were not correlated with self-reported correct dosing (ρ = -0.03, P = 0.68), nor with MEMS correct dosing (ρ = 0.06, P = 0.40). However, a weak significant negative correlation between D measures for implicit attitudes and disease activity scores was found (ρ = -0.17, P = 0.04), where more negative implicit attitudes (D measures) were associated with higher disease activity scores. No correlations were found between implicit health-related associations and any of the following variables: correct dosing measured with self-report (ρ = 0.04, P = 0.51), correct dosing measured with MEMS (ρ = -0.002, P = 0.97), and disease activity scores (ρ = -0.07, P = 0.42). [/fig]
[fig] Fig 2: Fig 2. Correlation between implicit attitudes (D measure) and explicit beliefs about medication (necessityconcerns differential scores). Higher NCD-scores indicate that necessity beliefs outweigh concern beliefs, whereas higher D measures for implicit attitudes indicate more positive attitudes towards cDMARDs (ρ = 0.13, P = 0.05). Abbreviations: cDMARD (conventional disease-modifying antirheumatic drug), NCD (necessity-concerns differential). [/fig]
[fig] Fig 3: The proportion of adherent patients with rheumatoid arthritis over time based on MEMS correct dosing adherence. Based on D measures and mean scale scores for explicit attitudes, patients were categorized in (in) congruent positive or negative attitudinal profiles. [/fig]
[fig] S1: File. Design, procedures and analyses of SC-IATs. (PDF) S2 File. Bipolar Evaluative Adjective Scale. (PDF) Table. List of abbreviations. (PDF) S2 [/fig]
[table] Table 1: Study sample characteristics.Categories were high versus medium-low for educational level, living alone versus living together (with children and/or partner) for residential status, and for ethnic background Dutch versus other. High educational level included a university degree or a degree in universities of applied sciences. P-values were calculated by Pearson chi-square tests or Two-sample t-tests. P-values �0.05 were considered statistically significant. Abbreviations: CQR (Compliance Questionnaire on Rheumatology); [/table]
[table] Table 2: Description of profiles of patients diagnosed with rheumatoid arthritis based on (in)congruent implicit and explicit attitudes. [/table]
[table] Table 3: Results nested linear regression models: Contribution of implicit attitudes over and above patient characteristics, disease characteristics, treatment characteristics and explicit measures in patients with rheumatoid arthritis. [/table]
[table] Table 4: Results nested linear regression models: contribution of implicit health-related associations over and above patient characteristics, disease characteristics, treatment characteristics and explicit measures in patients with rheumatoid arthritis. [/table]
|
Volcano-like Behavior of Au-Pd Core-shell Nanoparticles in the Selective Oxidation of Alcohols
Gold-palladium (AuPd) nanoparticles have shown significantly enhanced activity relative to monometallic Au and Pd catalysts. Knowledge of composition and metal domain distributions is crucial to understanding activity and selectivity, but these parameters are difficult to ascertain in catalytic experiments that have primarily been devoted to equimolar nanoparticles. Here, we report AuPd nanoparticles of varying Au:Pd molar ratios that were prepared by a seed growth method. The selective oxidation of benzyl alcohol was used as a model reaction to study catalytic activity and selectivity changes that occurred after varying the composition of Pd in bimetallic catalysts. We observed a remarkable increase in catalytic conversion when using a 1051 Au:Pd molar ratio. This composition corresponds to the amount of Pd necessary to cover the existing Au cores with a monolayer of Pd as a full-shell cluster. The key to increased catalytic activity derives from the balance between the number of active sites and the ease of product desorption. According to density functional theory calculations, both parameters are extremely sensitive to the Pd content resulting in the volcano-like activity observed.O ptimizing the architectures of bimetallic nanoparticles constitutes a major goal to merge synthesis with structure-activity relationships that allow the rational design of catalysts 1,2 . Gold-palladium nanoparticles (AuPd NPs) have shown significantly enhanced catalytic activity relative to monometallic Au and Pd NP catalysts in prototypical reactions, such as CO oxidation 3,4 , and in industrially challenging reactions, such as vinyl acetate monomer (VAM) synthesis 5 , the direct synthesis of hydrogen peroxide from H 2 and O 2 6 , alkene epoxidation 7 , and the oxidation of hydrocarbons 8 and alcohols 9,10 . Bimetallic nanoparticles are formed by metallic domain distributions ranging from homogeneous alloys (AB) to core-shell (A core @B shell or B core @A shell ) nanoparticles. The type of morphology prepared depends on the synthetic protocol, the miscibility of both metals, and post-synthesis treatment. Pd and Au are miscible in almost all compositions, forming AuPd alloy NPs 5,11-14 . The AuPd alloy NPs most frequently used in catalytic applications contain a 151 Au:Pd molar ratio. However, it has been well documented that the surface of a AuPd alloy NP differs from its corresponding bulk concentration15,16 . Enache et al. 9 have shown that Pd segregates towards the surface upon calcination, producing alloy NPs with a Pdrich shell and an Au-rich core. Therefore, the active catalyst has a surface that is significantly enriched with Pd, obtained by the metal's migration to the surface during the annealing process. Additionally, core-shell NPs can be produced by the reduction of Pd over preformed Au NPs, and vice versa. Henning et al 10 have shown the preparation of core-shell Au-Pd catalysts by the growth of Pd on icosahedral Au seeds through thermodynamic control. In catalytic studies, they identified a 12 atomic-layer-thick Pd shell (2.2 nm) as the optimum catalyst for benzyl alcohol oxidation. However, no mechanistic explanation was given for the observed composition effects. NPs with different Au:Pd molar ratios have also been suggested as the most active catalysts in other studies 8,17-21 . Thus, the question of which is the best Au:Pd ratio in the optimal catalyst for a given reaction and why, still remains unanswered. Here we synthesized Au core @Pd shell NPs with low Pd loading and investigated how the catalytic activity and selectivity were affected both experimentally and theoretically.Results and DiscussionThe preparation of bimetallic Au@Pd nanoparticles was based on the reduction of varying amounts of palladium over supported gold nanoparticles (Au NPs). The catalyst support chosen for this study is comprised of magnetic cores (Fe 3 O 4 , 10 nm) spherically coated with silica (spheres of ,40 nm in diameter)22. The catalyst support was loaded with Au 31 ions (0.9 wt%, determined by ICP OES), and supported Au NPs were formed by controlled OPEN SUBJECT AREAS: CATALYST SYNTHESIS NANOPARTICLES
reduction in molecular hydrogen. The morphology and size of the Au NPs used as seeds for the deposition of Pd were determined using a scanning transmission electron microscope (STEM). The Au NPs demonstrated a mean diameter of 12.0 6 3.2 nm, according to representative High Angle Annular Dark Field (HAADF)-STEM images and the size distribution histogram fitted to a Gaussian function shown in [fig_ref] Figure 1 |: Catalytic performance of the Au@Pd core-shell NP catalysts in the oxidation of... [/fig_ref]. In a second step, Pd (II) acetate solutions in benzyl alcohol containing variable amounts of Pd 21 ions were added over the Au cores and the reduction of Pd occurred in molecular hydrogen for 30 minutes at 100uC and 4 bar H 2 . The amount of Au (in mols) in the resultant Au@Pd NPs remained fixed, while the Pd composition varied from 1 to 40 mol % relative to Au, corresponding to 0 to 4 monolayers (ML) of Pd. The aerobic oxidation of benzyl alcohol was used as a model reaction to study the activity and selectivity of the Au@Pd core-shell NP catalysts [fig_ref] Figure 1 |: Catalytic performance of the Au@Pd core-shell NP catalysts in the oxidation of... [/fig_ref]. Conversion of benzyl alcohol and selectivity obtained by selected Au@Pd core-shell NP catalysts and monometallic Au and Pd NP catalysts are presented in [fig_ref] Table 1 |: Catalytic oxidation of benzyl alcohol by Au, Pd and Au@ Pd catalysts [/fig_ref]. The monometallic Au catalyst promoted high selectivity for benzaldehyde formation, but less than 2% conversion [fig_ref] Table 1 |: Catalytic oxidation of benzyl alcohol by Au, Pd and Au@ Pd catalysts [/fig_ref] entry 1). Both conversion and selectivity were poor for the monometallic Pd catalyst prepared by adding the corresponding amount of Pd to the support without gold [fig_ref] Table 1 |: Catalytic oxidation of benzyl alcohol by Au, Pd and Au@ Pd catalysts [/fig_ref]. The bimetallic catalysts exhibited a strong dependence between the activity and the amount of palladium added as shell metal. The highest benzyl alcohol conversion was achieved by adding 10 mol % Pd in respect to the existing amount of Au (i.e., Au:Pd 5 1051) [fig_ref] Table 1 |: Catalytic oxidation of benzyl alcohol by Au, Pd and Au@ Pd catalysts [/fig_ref] , entry 5). A sharp increase in benzyl alcohol conversion was observed by adding up to 10 mol % Pd (or 1 ML Pd), which was followed by a decrease as the Pd content was increased, as shown in [fig_ref] Figure 1 |: Catalytic performance of the Au@Pd core-shell NP catalysts in the oxidation of... [/fig_ref]. The selectivity to benzaldehyde was only slightly affected by the Pd content of Au@Pd catalysts. After each reaction, the catalysts were easily separated by placing a magnet on the external reactor wall due to the magnetic properties of the catalyst support. The most active catalyst (Au:Pd 5 1051) was recycled and reused in successive reactions, maintaining the high selectivity to benzaldehyde. Fluctuation in the reaction conversion was reported before for an alloyed AuPd catalyst [bib_ref] Magnetically recoverable AuPd nanoparticles prepared by a coordination capture method as a..., Silva [/bib_ref] and can be related to the presence of amino-groups on the catalysts support (used to improve the impregnation of gold on the support 24 ) and possible interactions between those groups and the reaction products.
The metal composition of the most active catalyst was determined by Flame Atomic Absorption Spectrometry (FAAS) to consist of mole fraction percentages of 89.9% Au and 9.1% Pd (i.e., Au:Pd 5 1051), which agrees with the nominal composition. This was also the case for other catalysts, as shown in [fig_ref] Table 3 |: Composition of Au and Pd in selected Au core @Pd shell catalysts... [/fig_ref] , which suggests deposition of the desired amount of palladium.
The most active catalyst was also investigated by X-ray energy dispersive spectroscopy in a scanning transmission electron microscope (XEDS-STEM) to determine both the compositional distribution between particles and the existence of compositional rich domains within each particle. The mean composition (mole fraction) of the catalyst particles was determined to be 10.2 6 6.7% Pd (see ESI for the details, [fig_ref] Figure 2 |: HAADF-STEM image of a supported catalyst particle Au [/fig_ref]. The distribution of the metallic components within each particle was investigated using X-ray energy dispersive spectroscopy spectrum imaging (XEDS-SI). The XEDS-SI technique enables direct visualization of the nanoscale structural and compositional features of the synthesized core-shell particles [bib_ref] Multimetallic Au/FePt 3 nanoparticles as highly durable electrocatalyst, Wang [/bib_ref]. [fig_ref] Figure 2 |: HAADF-STEM image of a supported catalyst particle Au [/fig_ref] shows a HAADF-STEM image of a quasi-spherical catalyst particle with the respective Au and Pd maps. In the compositional maps, the bright intensity in each pixel is a measurement of the number of X-ray peak counts at each corresponding position. Although the number of X-ray counts for Au La is larger than for Pd La in the investigated material [fig_ref] Figure 2 |: HAADF-STEM image of a supported catalyst particle Au [/fig_ref] , the brightness distribution was normalized in the maps shown in [fig_ref] Figure 2 |: HAADF-STEM image of a supported catalyst particle Au [/fig_ref]. In the Au map, X-ray counts are measured throughout the particle, resulting in a hemisphere-shaped line scan. A different situation is observed in the Pd distribution because the number of counts increases steeply at the borders of the particle and remains nearly constant throughout the remainder of the particle. This distribution shows that the particles investigated are morphologically structured with an Au-rich core and a Pd-rich shell [bib_ref] Electron energy loss spectroscopic investigation of polycrystalline Au, Pd and a Pd-Au..., Hagelin-Weaver [/bib_ref]. The less active catalyst with Au:Pd 5 1054 is also morphologically structured with an Au-rich core and a Pd-rich shell [fig_ref] Figure 3 |: The three model systems employed in the DFT calculations [/fig_ref]. Monometallic nanoparticles were not found in the investigated samples.
In the catalytic studies, we observed a volcano-like behavior characterized by a steady increase in benzyl alcohol conversion as the quantity of added Pd (shell metal) was increased relative to Au (core metal), with a consistent decrease in conversion for more than 10 mol % Pd. If the Pd atoms are primarily on the surface of Au, as observed in the morphological studies, the composition of the most active catalyst (89.9% Au and 9.1% Pd) correlates with the composition required for the complete coverage of Au cores (d 5 12.0 6 3.2 nm) with one atomic layer of Pd (see [fig_ref] Table 1 |: Catalytic oxidation of benzyl alcohol by Au, Pd and Au@ Pd catalysts [/fig_ref]. A special behaviour related to the monolayer regime was reported previously for other alloys [bib_ref] Surface strain versus substrate interaction in heteroepitaxial metal layers: Pt on Ru(0001), Schlapka [/bib_ref] [bib_ref] Ru-Pt core-shell nanoparticles for preferential oxidation of carbon monoxide in hydrogen, Alayoglu [/bib_ref] [bib_ref] Preferential CO oxidation in hydrogen: reactivity of core-shell nanoparticles, Nilekar [/bib_ref]. Theoretical simulations are fundamental to the understanding of the volcano-like behavior reported in [fig_ref] Figure 1 |: Catalytic performance of the Au@Pd core-shell NP catalysts in the oxidation of... [/fig_ref]. Density functional theory-based investigations of AuPd systems have concentrated on the structure 29-32 , oxygen activation 33 , or VAM synthesis [bib_ref] Segregation at the surface of an Au/Pd alloy exposed to CO, Soto-Verdugo [/bib_ref] [bib_ref] Monte Carlo and density functional theory analysis of the distribution of gold..., Boscoboinik [/bib_ref] [bib_ref] Coverage effects on the palladium-catalyzed synthesis of vinyl acetate: comparison between theory..., Calaza [/bib_ref] [bib_ref] Temperature and pressure effects in CO titration of ensembles in PdAu(111) alloys..., Garcia-Mota [/bib_ref] [bib_ref] Template effects in vinyl acetate synthesis on PdAu surface alloys: a density..., Garcia-Mota [/bib_ref]. Here we modelled the reaction process on three representative surfaces with an ordered surface alloy (ML denotes the number of Pd monolayers grown on the Au seeds): Au@Pd 3 Au, Au@Pd(1 ML) and Au@ Pd(2 ML) [fig_ref] Figure 3 |: The three model systems employed in the DFT calculations [/fig_ref] ; and Pd.
With these material models and the PBE functional [bib_ref] Generalized gradient approximation made simple, Perdew [/bib_ref] , we inspected the paths leading to the formation of the aldehyde, considering the following elementary steps:
(1) Alcohol adsorption: C 6 H 5 CH 2 OH(g)1* R C 6 H 5 CH 2 OH* (2) Oxygen adsorption:
[formula] O 2 (g) 1 2* R 2O* (3) First H transfer: C 6 H 5 CH 2 OH* 1 O* R C 6 H 5 CH 2 O* 1 OH* (4) Second H transfer: C 6 H 5 CH 2 O* 1 OH* R C 6 H 5 CHO* 1 H 2 O* (5) Water desorption: H 2 O* R H 2 O(g)1* (6) Aldehyde desorption: C 6 H 5 CHO* R C 6 H 5 CHO(g)1* (7) Unselective path: C 6 H 5 CHO* 1 H 2 O* R C 6 H 5 CHOH* 1 OH* [/formula]
The reaction profile is presented in [fig_ref] Figure 4 |: Reaction profile for the selective oxidation of benzyl alcohol to the corresponding... [/fig_ref] for the bimetallic systems only. The activation barriers, not shown in the figure, and all data for the Pd-only system, are presented in. Analysis of the results in [fig_ref] Figure 4 |: Reaction profile for the selective oxidation of benzyl alcohol to the corresponding... [/fig_ref] leads to the conclusion that adsorption of the alcohol is difficult (endothermic by 0.45 eV) on Au@Pd 3 Au, and thus the reaction cannot progress adequately in such structures because the number of molecules adsorbed on the surface is small. Increasing the amount of Pd on the surface improves adsorption rendering it exothermic (20.31 eV) as observed for Au@ Pd(1 ML). When additional Pd is present in the slab, the binding energy is very strong, 21.69 eV for Au@Pd(2 ML). The reaction has easily attainable energy barriers for all of the steps (see. Water desorption (step 5) is exothermic for systems with Pd shells up to one monolayer (20.66 and 20.22 eV for Au@Pd 3 Au and Au@ Pd(1 ML), respectively), and endothermic for NPs having higher Pd content. In the reaction mechanism, the OH intermediate is a short living species due to the difficulties found in its formation and the easiness at which the next reaction takes place. Therefore, the main path for the unselective process is the back transfer of one proton from water, i.e. step 7. The reduced energy barrier for water desorption guarantees the selectivity towards benzaldehyde because the barrier for this step is, in all cases, lower than H transfer from water to the aldehyde. The final aldehyde product is easily desorbed from the Au@Pd 3 Au model, but its binding energy is large on Pd monolayers: 0.8 and 1.1 eV, for 1 and 2 ML, respectively. As for the activity, reactant adsorption (1.7 vs. 0.3 eV) and product desorption (1.1 vs. 0.8 eV) are far two strong on the two monolayer system compared to the single layer to ensure a rapid recycling of the clean surface.
The computational approach is in line with the mechanistic investigations in the literature. For instance, Enache et al. [bib_ref] Solvent-free oxidation of primary alcohols to aldehydes using Au-Pd/TiO 2 catalysts, Enache [/bib_ref] found zero order for oxygen, which is in agreement with the ease of dissociation of oxygen reported in previous calculations [bib_ref] The role of long-lived oxygen precursors on AuM alloys (M 5 Ni,..., Garcia-Mota [/bib_ref]. In addition, the apparent activation energy reported was 0.47 eV. This value is similar to. However, the strong Pd-Au ligand effect is observed and indeed the ligand contribution affects alcohol adsorption to such an extension that the binding is 20.31 eV only. The calculations including entropy and dispersion were done as follows: DG 5 DE 2 TDS; where S was only considered for the gas-phase species and DE included the dispersion terms calculated through the dispersion model introduced by Grimme et al. [bib_ref] A consistent and accurate ab initio parametrization of density functional dispersion correction..., Grimme [/bib_ref] and the metal parameters from Tkatchenko et al [bib_ref] Density-functional theory with screened van der waals interactions for the modeling of..., Ruiz [/bib_ref]. In the case of Pd, to analyze unselectivity, water desorption and unselective H transfer need to be considered. If entropic and dispersion contributions are considered desorption is endothermic by 0.3 eV while the barrier for proton transfer is as low as 0.14 eV, therefore selectivity is compromised. The analysis of the reaction network accounts for the observation of a maximum in the experimental catalytic results. On the left side of the volcano-like plot [fig_ref] Figure 1 |: Catalytic performance of the Au@Pd core-shell NP catalysts in the oxidation of... [/fig_ref] corresponding to low Pd loadings, DFT results indicate that adsorption of the alcohol is controlling the activity. The electronic properties of AuPd alloys have been extensively analyzed in the literature [bib_ref] Segregation at the surface of an Au/Pd alloy exposed to CO, Soto-Verdugo [/bib_ref] [bib_ref] Monte Carlo and density functional theory analysis of the distribution of gold..., Boscoboinik [/bib_ref] [bib_ref] Coverage effects on the palladium-catalyzed synthesis of vinyl acetate: comparison between theory..., Calaza [/bib_ref] [bib_ref] Temperature and pressure effects in CO titration of ensembles in PdAu(111) alloys..., Garcia-Mota [/bib_ref]. Actually, only sites with no Au would be active for benzyl alcohol adsorption, and thus, activity is proportional to the number of Pd ensembles (i.e., at least 6 Pd atoms on the surface) that can accommodate benzyl alcohol. However, increasing benzyl alcohol adsorption subsequently increases the desorption energy required to clean the surface from the aldehyde product. This is a drawback when having more than 1 ML of Pd, as the binding energy of the aldehyde is far too large to remove the product from the surface and leave clean ensembles for the next catalytic cycle. As a consequence, the best balance is obtained when 1 ML of Pd covers the Au nanoparticles because benzyl alcohol adsorption can then occur everywhere and desorption is still feasible. For the Pd-only system, the barrier for the second H transfer is the most energy-demanding step, 1.32 eV, and thus implies that the reaction is slow at the temperatures at which AuPd alloys are active. In addition, once the aldehyde is formed, the water desorption energy and the barrier for H stripping are comparable (0.10 vs. 0.14 eV, respectively), and therefore, compromising selectivity determining step is not as favorable as for the alloys. When extending the study to other unsaturated alcohols the selectivity controlled by the competition between steps (5) and (7) would follow the same general rule. As for the activity, the adsorption of these unsaturated alcohols (like crotyl alcohol) needs smaller ensembles than those of the corresponding benzyl alcohol, but again the number of sites would maximize for a composition of 1 ML of Pd on top of the Au core. Therefore, the present results can be extrapolated to other unsaturated alcohols.
# Conclusions
We have shown that the rational design based on the density functional theory and the controlled synthesis of Au@Pd core-shell catalysts with specific compositions and metal domain distributions can be used as a strategy to enhance the catalytic properties of Au while maintaining its high selectivity for the oxidation of benzyl alcohol to benzaldehyde. This finding shows that many opportunities remain to obtain enhanced properties in bimetallic nanomaterials by tailoring the amount of each metal and the metal domain distribution, and opens new possibilities in the field of nanomaterials synthesis and design.
# Methods
Synthesis of the catalyst support. The core-shell Fe 3 O 4 @SiO 2 nanocomposite was prepared and modified with (3-aminopropyl)triethoxysilane as reported in the literature [bib_ref] Recoverable rhodium nanoparticles: synthesis, characterization and catalytic performance in hydrogenation reactions, Jacinto [/bib_ref].
Synthesis of the supported Au monometallic catalyst. The catalyst support, silicacoated magnetite nanoparticles previously modified with (3aminopropyl)triethoxysilane (500 mg), was added to an aqueous solution of HAuCl 4 (40 mL, 0.215 g L 21 ) with pH 6. The mixture was stirred at 25uC for 2 h. The solid was then collected magnetically from the solution and washed twice with hot distilled water and twice with ethanol. The reduction step was performed in a Fischer-Porter glass reactor pressurized with hydrogen gas. The gold precursor was dispersed in cyclohexane and heated to 100uC under H 2 at 4 bar. The mixture was stirred for 2 h, and the color of the solid changed from brown to dark red.
Synthesis of the supported AuPd bimetallic catalysts. The preparation of bimetallic catalysts consisted of an addition of the desired amount of a Pd(OAc) 2 solution in benzyl alcohol (20 to 800 mL, [Pd 21 ] 5 1.68 3 10 23 mol L 21 ) to 75 mg of Au monometallic catalyst (3.4 mmol Au), which was then added to a Fischer-Porter 100 mL glass reactor. Benzyl alcohol was added to achieve a total volume of 1.0 mL and the reactor was pressurized with hydrogen gas at 4 bar. The mixture was stirred for 30 min at 100uC. The color of the solid changed to black. The reactor was cooled to room temperature, the hydrogen gas was relieved, and the reaction mixture was kept inside the reactor for catalytic studies.
Catalytic experiments. All reactions were performed using a modified Fischer-Porter 100 mL glass reactor. In a typical reaction, the glass reactor was loaded with the supported Au catalyst (75 mg, 3.4 mmol Au) and benzyl alcohol (9.6 mmol), or the bimetallic Au-Pd catalyst-containing mixture was prepared as described above. The reactor was purged three times with O 2 , leaving the vessel at 6 bar. The temperature was maintained at 100uC with an oil bath on a hot stirring plate connected to a digital controller (ETS-D5 IKA). The reactions were constantly stirred using Teflon-coated magnetic stir bars for the desired amount of time. The catalyst was recovered magnetically by affixing a magnet onto the reactor wall. The products were collected with a syringe and analyzed by gas chromatography (GC) and gas chromatography-mass spectrometry (GC-MS).
Morphological investigation. Catalyst samples were prepared by sonicating the catalyst powder in isopropanol. A drop of the resulting dispersion was placed on a thin carbon film, which was deposited on standard 400-mesh TEM copper grids and air-dried. The size distribution of the supported Au core particles was investigated by HAADF-STEM using a JEOL-JEM 2100F TEM microscope available at the LNNano Laboratory (CNPEM, Campinas, Brazil). The mean diameter distribution was measured by HAADF image analysis using Image-Pro Plus 6.0 software. Two hundred nanoparticles were measured to build the size distribution histogram that is presented. Compositional distribution was measured on the same instrument by acquiring individual XEDS-STEM spectra of a sample of 62 supported particles. Particle compositional maps were acquired by XED-Spectrum Imaging with a Digital Micrograph 1.8 system (Gatan Inc.) controlling a Thermo-Noran XEDS. The SIs were acquired at a 15 3 15 pixel resolution with a dwell time of 2 s/pixel, using the drift correction facility every 100 s. XED spectra were acquired in the 0-10 keV energy range at each pixel. Electron probe sizes of approximately 0.7 nm were obtained by operating the JEM 2100F instrument in STEM mode, allowing enough current [fig_ref] Figure 4 |: Reaction profile for the selective oxidation of benzyl alcohol to the corresponding... [/fig_ref].
www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 5766 | DOI: 10.1038/srep05766 density at each sample point to acquire statistically significant X-ray counts for the investigated elements.
Computational Methods. Density Functional Theory was applied to slabs modeling the system. The VASP program was employed to this end [bib_ref] Efficiency of ab-initio total energy calculations for metals and semiconductors using a..., Kresse [/bib_ref] [bib_ref] Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis..., Kresse [/bib_ref]. The functional of choice was PBE [bib_ref] Generalized gradient approximation made simple, Perdew [/bib_ref]. Inner electrons were replaced by frozen cores PAW [bib_ref] Projector augmented-wave method, Blochl [/bib_ref] and valence electrons were expanded in plane waves with cut-off energy of 450 eV. As the particles are large enough only (111) surfaces, the most stable for fcc metals, have been employed in the simulations. Three different models were employed to represent the surface of the Pd grown on Au particles: the first one is a partial monolayer on the surface of Au@Pd 3 Au composition, this is following one of the ordered alloys, a single Pd monolayer on Au, and a Pd bilayer on Au (see [fig_ref] Figure 3 |: The three model systems employed in the DFT calculations [/fig_ref] and Pd alone. The slabs contain four layers of metal in total and in all cases overlayers follow the lattice parameter of Au. The vacuum space interleaving the slabs is at 15 Å . As the adsorbates are large we have employed a surface reconstructions p(6 3 4) that can account for the adsorption of the benzyl alcohol and its derivates. On these supercells the k-point sampling employed are 2 3 5 3 1 Monkhorst-Pack points centered at gamma point [bib_ref] Special points for Brillouin-zone integrations, Monkhorst [/bib_ref]. Transition state searches were performed with the climbing image version of the Nudged Elastic Band method [bib_ref] A climbing image nudged elastic band method for finding saddle points and..., Henkelman [/bib_ref].
[fig] Figure 1 |: Catalytic performance of the Au@Pd core-shell NP catalysts in the oxidation of benzyl alcohol. The amount of Au is fixed (3.4 mmol), while the amount of Pd varies from 0 to 40 mol % (i.e., 0 to 1.4 mmol). [/fig]
[fig] Figure 2 |: HAADF-STEM image of a supported catalyst particle Au:Pd 5 1051 and the corresponding Au and Pd maps. The particle composition distribution is observed in the line scans, measured from the regions delimited by the lines indicated in both maps. [/fig]
[fig] Figure 3 |: The three model systems employed in the DFT calculations.Left a Pd 3 Au overlayer on Au; center a Pd monolayer on Au and right two Pd monolayers on Au. Blue spheres stand for Pd atoms and yellow (olive) for Au. [/fig]
[fig] Figure 4 |: Reaction profile for the selective oxidation of benzyl alcohol to the corresponding aldehyde. The insets schematically represent the different steps in the reaction for the Au@Pd(1 ML) model: blue spheres Pd, yellow Au, grey C, red O, white H (see also [/fig]
[table] Table 1 |: Catalytic oxidation of benzyl alcohol by Au, Pd and Au@ Pd catalysts [/table]
[table] Table 3 |: Composition of Au and Pd in selected Au core @Pd shell catalysts determined by FAAS [/table]
|
PScL-2LSAESM: bioimage-based prediction of protein subcellular localization by integrating heterogeneous features with the two-level SAE-SM and mean ensemble method
Motivation: Over the past decades, a variety of in silico methods have been developed to predict protein subcellular localization within cells. However, a common and major challenge in the design and development of such methods is how to effectively utilize the heterogeneous feature sets extracted from bioimages. In this regards, limited efforts have been undertaken. Results: We propose a new two-level stacked autoencoder network (termed 2L-SAE-SM) to improve its performance by integrating the heterogeneous feature sets. In particular, in the first level of 2L-SAE-SM, each optimal heterogeneous feature set is fed to train our designed stacked autoencoder network (SAE-SM). All the trained SAE-SMs in the first level can output the decision sets based on their respective optimal heterogeneous feature sets, known as 'intermediate decision' sets. Such intermediate decision sets are then ensembled using the mean ensemble method to generate the 'intermediate feature' set for the second-level SAE-SM. Using the proposed framework, we further develop a novel predictor, referred to as PScL-2LSAESM, to characterize image-based protein subcellular localization. Extensive benchmarking experiments on the latest benchmark training and independent test datasets collected from the human protein atlas databank demonstrate the effectiveness of the proposed 2L-SAE-SM framework for the integration of heterogeneous feature sets. Moreover, performance comparison of the proposed PScL-2LSAESM with current state-of-the-art methods further illustrates that PScL-2LSAESM clearly outperforms the existing stateof-the-art methods for the task of protein subcellular localization. Availability and implementation: https://
# Introduction
The knowledge regarding the precise subcellular location of a protein is crucial for the determination of its function and involved biological processes [bib_ref] Hum-mPLoc 3.0: prediction enhancement of human protein subcellular localization through modeling the..., Zhou [/bib_ref]. Location proteomics is concerned with the large-scale study of protein localization inside a cell [bib_ref] Location proteomics: a systems approach to subcellular location, Murphy [/bib_ref] [bib_ref] Location proteomics: systematic determination of protein subcellular location, Newberg [/bib_ref]. In location proteomics, various methods are used to analyze and predict protein subcellular localization such as wet-lab experiments or computational methods. Determining the protein subcellular localization via wet-lab experiments is often time-consuming and labor-intensive. Due to the importance of subcellular location of proteins and as a useful alternative to facilitate experimental characterization of protein subcellular localization, automatic computational methods are attracting a great deal of interest in recent years, representing the main focus in location proteomics.
Over the past decades, a great variety of computational methods have emerged for characterization of protein subcellular localization from diverse protein data sources, including amino acid sequences or bioimages. Based on the data sources, these methods can be generally categorized into either one-dimensional sequence-based or two-dimensional (2D) image-based methods which can be further grouped into single-label or multi-label methods [bib_ref] pLoc-mHum: predict subcellular localization of multi-location human proteins via general PseAAC to..., Cheng [/bib_ref] [bib_ref] Incorporating label correlations into deep neural networks to classify protein subcellular location..., Hu [/bib_ref] [bib_ref] SLPred: a multi-view subcellular localization prediction tool for multi-location human proteins, Ö Zsarı [/bib_ref] [bib_ref] Bioimage-based protein subcellular location prediction: a comprehensive review, Xu [/bib_ref] [bib_ref] Learning complex subcellular distribution patterns of proteins via analysis of immunohistochemistry images, Xu [/bib_ref] [bib_ref] Automated classification of protein subcellular localization in immunohistochemistry images to reveal biomarkers..., Xue [/bib_ref] [bib_ref] Image-based classification of protein subcellular location patterns in human reproductive tissue by..., Yang [/bib_ref] [bib_ref] SubMito-XGBoost: predicting protein submitochondrial localization by fusing multiple feature information and eXtreme..., Yu [/bib_ref] [bib_ref] Accurate prediction of multi-label protein subcellular localization through multi-view feature learning with..., Zhang [/bib_ref]. In the case of sequence-based methods, they can effectively determine the location of the protein [bib_ref] Cell-PLoc: a package of web servers for predicting subcellular localization of proteins..., Chou [/bib_ref] [bib_ref] Human protein subcellular localization with integrated source and multi-label ensemble classifier, Guo [/bib_ref] and protein properties [bib_ref] BioSeq-BLM: a platform for analyzing DNA, RNA and protein sequences based on..., Li [/bib_ref] [bib_ref] BioSeq-Analysis: a platform for DNA, RNA and protein sequence analysis based on..., Liu [/bib_ref] [bib_ref] PTM-ssMP: a web server for predicting different types of post-translational modification sites..., Liu [/bib_ref] [bib_ref] DeepPhos: prediction of protein phosphorylation sites with deep learning, Luo [/bib_ref]. However, as amino acid sequences do not change whenever the translocation takes place, this will make them unfit for the detection of subcellular translocation of proteins. In this regard, the image-based methods can serve as an alternative complementary to the sequence-based methods because they can unravel the crucial information regarding the spatial distribution of proteins across the normal and cancerous tissues, as well as their location changes in various tissues. Therefore, development of 2D image-based computational methods to facilitate the identification of protein subcellular localization has become an increasingly important problem in bioinformatics and computational biology [bib_ref] Bioimage informatics: a new category in bioinformatics, Peng [/bib_ref].
During the development of image-based computational methods for analyzing protein subcellular localization, considerable challenges often exist for statistical and machine learning models, especially related to feature extraction, feature selection (FS), feature integration and classification. For example, DNA distribution and Haralick texture features, which belong to subcellular location features (SLFs) [bib_ref] A neural network classifier capable of recognizing the patterns of all major..., Boland [/bib_ref] , are frequently utilized to represent and encode the global information from the bioimages. In addition, a variety of local features such as local binary pattern (LBP) [bib_ref] Multiresolution gray-scale and rotation invariant texture classification with local binary patterns, Ojala [/bib_ref] , completed local binary pattern (CLBP) [bib_ref] A completed modeling of local binary pattern operator for texture classification, Guo [/bib_ref] , local ternary pattern, local quinary pattern [bib_ref] Local binary patterns variants as texture descriptors for medical image analysis, Nanni [/bib_ref] , rotation invariant cooccurrence among adjacent local binary patterns (RICLBP) [bib_ref] Rotation invariant co-occurrence among adjacent LBPs, Nosaka [/bib_ref] and locally encoded transform feature histogram (LETRIST) [bib_ref] LETRIST: locally encoded transform feature histogram for rotation-invariant texture classification, Song [/bib_ref] are utilized to extract the local micropatterns from images. Current studies have shown that extracting both global and local features can help improve the predictive capabilities of the developed methods [bib_ref] An image-based multi-label human protein subcellular localization predictor (iLocator) reveals protein mislocalizations..., Xu [/bib_ref] [bib_ref] Incorporating organelle correlations into semi-supervised learning for protein subcellular localization prediction, Xu [/bib_ref] [bib_ref] MIC_locator: a novel image-based protein subcellular location multi-label prediction model based on..., Yang [/bib_ref]. Similarly, at the FS stage, a number of studies have proposed different FS algorithms to effectively select the optimal features from the extracted features [bib_ref] Bioimage-based prediction of protein subcellular location in human tissue with ensemble features..., Liu [/bib_ref] [bib_ref] A framework for the automated analysis of subcellular patterns in human protein..., Newberg [/bib_ref] [bib_ref] An organelle correlation-guided feature selection approach for classifying multi-label subcellular bio-images, Shao [/bib_ref] [bib_ref] PScL-HDeep: image-based prediction of protein subcellular location in human tissue using ensemble..., Ullah [/bib_ref]. Among such FS algorithms, the stepwise discriminant analysis (SDA) algorithmhas been widely adopted and shown to be effective for FS.
Use of multiple heterogeneous features is a common step during the development of automated models as it remains a challenging task to represent the global and local features from images based on single traditional handcrafted feature sets. As such, the difficulty in feature integration arises when multiple features are used to represent the protein image. One useful way to utilize multiple feature sets is to simply concatenate all the feature sets in a simple serial fashion. Many protein subcellular localization prediction methods have utilized this simple serial integration strategy to integrate all the feature sets and subsequently develop a predictor [bib_ref] Subcellular localization using fluorescence imagery: utilizing ensemble classification with diverse feature extraction..., Tahir [/bib_ref] [bib_ref] PScL-DDCFPred: an ensemble deep learning-based approach for characterizing multiclass subcellular localization of..., Ullah [/bib_ref] [bib_ref] Incorporating organelle correlations into semi-supervised learning for protein subcellular localization prediction, Xu [/bib_ref] [bib_ref] Bioimage-based protein subcellular location prediction: a comprehensive review, Xu [/bib_ref]. More recently, several studies have investigated new techniques other than simple serial integration [bib_ref] Human cell structure-driven model construction for predicting protein subcellular location from biological..., Shao [/bib_ref] [bib_ref] Learning protein subcellular localization multi-view patterns from heterogeneous data of imaging, sequence..., Wang [/bib_ref]. However, less attention is being paid to integrating multiple feature sets and accordingly, there still remains significant challenges as to how these multiple feature sets can be efficiently integrated.
From the algorithmic perspective, a number of different classification models have been developed to predict protein subcellular localization. For example, support vector machine [bib_ref] Support-vector networks, Cortes [/bib_ref] , error correcting output coding [bib_ref] Solving multiclass learning problems via error-correcting output codes, Dietterich [/bib_ref] , discriminant error correcting output coding [bib_ref] Discriminant ECOC: a heuristic method for application dependent design of error correcting..., Pujol [/bib_ref] , random forest (RF) [bib_ref] Random forests, Breiman [/bib_ref] and deep learning-based models [bib_ref] A fast learning algorithm for deep belief nets, Hinton [/bib_ref] [bib_ref] Long short-term memory, Hochreiter [/bib_ref] [bib_ref] Very deep convolutional networks for large-scale image recognition, Simonyan [/bib_ref] have been utilized efficiently. Despite the extensive efforts being undertaken, currently available computational approaches continue to have insufficient and limited performance of protein subcellular localization. This is particularly the case in terms of the overall success rate and as a consequence, there remains an exigent need to develop novel and high-performance predictors.
Motivated by the issues mentioned above, in this study, we make the following contributions in order to improve the predictive performance of protein subcellular localization: first, to ensure that the dataset is up to date and no mistakenly labeled data are included, we collect the high-quality datasets from the latest version of human protein atlas (HPA) databank [bib_ref] Tissue-based map of the human proteome, Uhlén [/bib_ref] as the collection of the latest datasets is highly desirable for the development of accurate predictors; Second, we design and develop a new classifier called the stacked autoencoder-SoftMax (SAE-SM) network; Third, using the designed SAE-SM, we further develop a two-level SAE-SM (2L-SAE-SM) framework based on the integration of multiple feature sets, and fourth, based on 2L-SAE-SM, we implement a bioimage-based protein subcellular localization predictor termed PScL-2LSAESM. Benchmarking results on the stringent 10-fold cross-validation using the benchmark training dataset and the independent test using the independent test dataset illustrate the effectiveness of the proposed framework.
# Materials and methods
## Benchmark datasets
The benchmark image datasets used in this study were collected from the publicly available Tissue Atlas of the HPA database (version 21, http://proteinatlas.org) [bib_ref] Towards a knowledge-based human protein atlas, Uhlen [/bib_ref] [bib_ref] Tissue-based map of the human proteome, Uhlén [/bib_ref]. The same criteria (i.e. reliability and validation scores) as [bib_ref] PScL-HDeep: image-based prediction of protein subcellular location in human tissue using ensemble..., Ullah [/bib_ref] were considered during the collection of protein entries in our datasets. The immunohistochemistry (IHC)-based brightfield microscopic images of these proteins were collected in this study. All the images belong to normal human tissues; according to the annotations in HPA, each IHC image in the benchmark datasets was labeled as one of the seven major subcellular location classes including cytoplasm (Cytopl.), endoplasmic reticulum (ER), Golgi apparatus (Gol.), mitochondrion (Mito.), lysosome (Lyso.), nucleus (Nucl.) and vesicles (Vesi.).
The benchmark training dataset, referred to as PScL2708, encompasses 2708 IHC images belonging to 23 different proteins. The subcellular location classes including cytoplasm, endoplasmic reticulum, Golgi apparatus, mitochondrion, lysosome, nucleus and vesicles contain 4, 3, 3, 3, 2, 4 and 4 different proteins, respectively. Similarly, we also collected the independent test dataset called PScL227 in our study. PScL227 has 227 IHC images belonging to seven distinguished proteins. Each protein belongs to one subcellular location class. [fig_ref] Table 1: Statistical distribution of the images across each subcellular location class [/fig_ref] summarizes the statistical distribution of the images across each subcellular location class for both the PScL2708 and PScL227 datasets.
## Image separation and feature extraction
All IHC bioimages in the HPA database are the mixture of DNA and protein stains. As we were interested in the subcellular localizations of proteins only, therefore, we first used the linear spectral separation (LIN) method (see Supplementary Text S1 for details regarding the linear spectral separation) to separate each original IHC image into DNA and protein channels. Next, heterogeneous features extracted from multiple aspects might reveal hidden information from protein image samples, which is useful for predicting protein subcellular localization. Therefore, we extracted various global and local heterogeneous features considering that the global and local features are expected to extract complementary information from protein images [bib_ref] Image-based classification of protein subcellular location patterns in human reproductive tissue by..., Yang [/bib_ref]. In our study, we extracted five types of heterogeneous feature sets from each IHC image. These included SLFs, LBP, CLBP, LETRIST and RICLBP with the dimensionalities of 840, 256, 906, 413 and 408, respectively. Previous studies have shown these features to be very effective in this field [bib_ref] PScL-HDeep: image-based prediction of protein subcellular location in human tissue using ensemble..., Ullah [/bib_ref] [bib_ref] PScL-DDCFPred: an ensemble deep learning-based approach for characterizing multiclass subcellular localization of..., Ullah [/bib_ref] [bib_ref] Incorporating organelle correlations into semi-supervised learning for protein subcellular localization prediction, Xu [/bib_ref]. SLFs are the global features which includes 4-dimensional DNA distribution and 836-dimensional Haralick texture features [bib_ref] A neural network classifier capable of recognizing the patterns of all major..., Boland [/bib_ref] [bib_ref] PScL-DDCFPred: an ensemble deep learning-based approach for characterizing multiclass subcellular localization of..., Ullah [/bib_ref] [bib_ref] An image-based multi-label human protein subcellular localization predictor (iLocator) reveals protein mislocalizations..., Xu [/bib_ref]. SLFs are very useful for extracting the global texture information from images [bib_ref] A neural network classifier capable of recognizing the patterns of all major..., Boland [/bib_ref]. The corresponding LBP features were extracted to characterize the local texture structure and detect micropatterns such as spots, edges and flat areas. In addition, CLBP and RICLBP features were also extracted to ensure the rotation invariance and information neglected by LBP. Similarly, LETRIST features were extracted because they could explicitly encode the joint information within the IHC image across the feature and scale spaces. For the sake of convenience, we termed LETRIST as LET in this study. A detailed description of SLFs, LBP, CLBP, LET and RICLBP is provided in Supplementary Text S2. In the current study, we accordingly named these five extracted heterogeneous feature sets as SLFs-Raw, LBP-Raw, CLBP-Raw, LET-Raw and RICLBP-raw, respectively.
## Feature selection
In our study, all the five extracted heterogeneous features (i.e. SLFs-Raw, LBP-Raw, CLBP-Raw, LET-Raw and RICLBP-raw) have high dimensionalities and as such, there might exist irrelevant, redundant and noisy information which may either cause overfitting or underfitting. In order to avoid dimension explosion and remove feature redundancy, the original extracted features need to be reduced by some FS algorithms. A series of studies utilized various FS algorithms during the analysis of protein subcellular localization; however, among all these algorithms, the SDA has proven to be more effective. Therefore, in this study, we also employed the SDA algorithm on each feature set.
For a given training dataset X ¼ fðx j ; y j Þg N j¼1 , where x j is the j-th image sample, y j is its corresponding label and N is the total number of features, let S t ¼ fðx j;t ; y j Þg N j¼1 be the t-th ð1 t TÞ heterogeneous feature set extracted from X, where x j;t is the feature vector extracted from the j-th image sample for the t-th heterogeneous feature set. Suppose that for each feature vector j (i.e. x j;t ) in the t-th heterogeneous feature set S t , ðx opt j;t ; y i Þ be its j-th data pair representing the j-th image sample of X, where x opt j;t 2 R d opt t is the optimal feature vector for the t-th optimal heterogeneous feature set, d opt t is the dimension of the optimal feature vector. For Nnumber of features, a corresponding t-th optimal heterogeneous feature set, denoted as S opt t ¼ fðx opt j;t ; y j Þg N j¼1 can be generated. For T heterogeneous feature sets ði:e:
[formula] S 1 ¼ fðx j;1 ; y j Þg N j¼1 ; S 2 ¼ fðx j;2 ; y j Þg N j¼1 ; . . . ; S t ¼ fðx j;t ; y j Þg N j¼1 ; . . . ; S T ¼ fðx j;T ; y j Þg N j¼1 Þ, a [/formula]
total of T optimal heterogeneous feature sets ði:e: S opt
[formula] 1 ¼ fðx opt j;1 ; y j Þg N j¼1 ; S opt 2 ¼ fðx opt j;2 ; y j Þg N j¼1 ; . . . ; S opt t ¼ fðx opt j;t ; y j Þg N j¼1 ; . . . ; S opt T ¼ fðx opt j;T ; y j Þg N j¼1 [/formula]
Þ can be selected. For more theoretical and mathematical details of SDA algorithms, please refer to the Supplementary Text S3.
In our study, we represent the optimal heterogeneous feature set of SLFs-Raw as SLFs-optimal, LBP-Raw as LBP-Optimal, CLBP-Raw as CLBP-Optimal, LET-Raw as LET-Optimal and RICLBPraw as RICLBP-Optimal, respectively.
## Stacked autoencoder
A single autoencoder (AE) [bib_ref] Learning internal representations by error propagation, Rumelhart [/bib_ref] consists of an input layer, a hidden layer and an output layer (for details about autoencoder, please refer to the Supplementary Text S4). In order to construct a stacked autoencoder (SAE), multiple AEs are stacked on top of each other. In other words, an SAE is a neural network consisting of multiple layers of AEs where the activation output features of the k-th hidden layer of AE are sent as an input to the (k þ 1)-th hidden layer of AE. In cases where an SAE is used as a classifier, a classification layer must be added as the top layer to compute and output the probabilities of the classes. The purpose of stacking multiple AEs is to boost the performance of the model.
In the current study, we used two encoders and the SoftMax (SM) activation function as a classification layer to construct our stacked autoencoder network, referred to as SAE-SM as shown in [fig_ref] Figure 1: Illustration of the SAE-SM-model architecture Text S6 provides a detailed description of... [/fig_ref]. Given the input x (Input Layer), the first encoder (Encoder 1) produces first hidden layer activation output features h ð1Þ (Features 1). The activation output features h ð1Þ will be then fed to the second encoder (i.e. Encoder 2) generating the second hidden layer activation output features h ð2Þ (i.e. Features 2). The activation output features h ð2Þ are finally fed to the SM classifier layer (SoftMax Classifier) to output the corresponding class probabilities. Each hidden layer size and hyperparameters are provided in Supplementary Text S5. In the hidden layers, we used the sigmoid activation function and also imposed the sparsity constraint on hidden units. Two training phases are involved during the training of our SAE-SM: (i) layer-by-layer pre-training which uses the unsupervised learning method and (ii) fine-tuning which uses the supervised back propagation (BP) method. For example, once the first AE is pre-trained based on the input x, the output (h ð1Þ ) of the first AE can then be input to the next AE. This procedure continues until the pretraining is accomplished (i.e. layer-by-layer pre-training). Finally, the pre-trained SAE-SM is fine-tuned using the BP algorithm (i.e. fine-tuning).
## Proposed 2l-sae-sm
In order to effectively integrate multiple feature sets, we propose the 2L-SAE-SM framework. [fig_ref] Figure 2: The architecture of the proposed 2L-SAE-SM model for integrating heterogeneous feature sets [/fig_ref] illustrates its architecture. As can be seen, 2L-SAE-SM is a two-level model where in the first level, T number of SAE-SMs, denoted as SAE-SM 1 , SAE-SM 2 , . . ., SAE-SM T-1 , SAE-SM T , are trained on the T optimal heterogeneous feature sets selected from the T raw heterogeneous feature sets via the SDA algorithm detailed in the Section 2.3 to further learn the hidden information from the corresponding optimal heterogeneous feature set. The Mean Ensemble (ME) method is applied in the middle of two levels to ensemble the outputs of the trained first-level SAE-SMs, whose output would be fed into the second-level SAE-SM ME for making the prediction.
Next, we describe the ME in detail below: Let fp t g T t¼1 be the T 'intermediate decision' sets, where p t denotes the t-th 'intermediate decision' set. Then the ME can be represented as:
[formula] F ME ¼ 1 T X T t¼1 p t(1) [/formula]
where F ME denotes the 'intermediate feature' set. A major challenge is how to train a 2L-SAE-SM on a given training dataset. In order to efficiently handle this issue, Supplementary Based on the 2L-SAE-SM framework, we developed a novel predictor PScL-2LSAESM to characterize image-based protein subcellular localization. An overview of the working flow of the proposed PScL-2LSAESM is illustrated in [fig_ref] Figure 3: Illustration of the architecture of the proposed PScL-2LSAESM model consistently achieved the... [/fig_ref]. Each major stage of the PScL-2LSAESM is described in Supplementary Text S7 and the system configuration settings are discussed in Supplementary Text S8.
## Evaluation indices
In this study, five commonly used performance indices specially designed for evaluating the performance of multiclass learning are employed. These included the overall accuracy (OA), Macroaverage Precision (Prec M ), Macroaverage Recall (Rec M ), Macroaverage F1-Score (F1-Score M ) and Matthews' Correlation Coefficient (MCC). In addition to these indices, the mean of the area under the receiveroperating characteristic (ROC) curves (AUC) denoted as meanAUC, the mean of area under precision-recall (PR) curves (AUPR) denoted as meanAUPR and the standard deviation of AUC and AUPR denoted as stdAUC and stdAUPR, respectively, are also used as the other four evaluation indices. All the performance evaluation indices are described in detail in the Supplementary Text S9.
Stringent k-fold cross-validation and independent validation tests are conducted to evaluate the performance of the proposed model. When performing k-fold cross-validation, the value of k was set to 10. It is noteworthy that when evaluating the performance of the model via stringent k-fold cross-validation, the features were selected independently in each fold of the train dataset to avoid biased evaluation of the model performance.
# Results
## Performance comparison of the extracted heterogeneous feature sets
In this section, we examined the discriminative capabilities of SLFs-Raw, LBP-Raw, CLBP-Raw, RICLBP-Raw and LET-Raw heterogeneous feature sets. The performance of each heterogeneous feature set was evaluated by performing 10-fold cross-validation on the benchmark training dataset PScL2708 with our designed SAE-SM classifier. The performance comparison of all the five heterogeneous feature sets in terms of OA, Rec M , Prec M , F1-Score M and MCC is provided in [fig_ref] Table 2: Performance comparison of the raw feature sets on 10fold cross-validation using the... [/fig_ref].
Several observations can be derived from [fig_ref] Table 2: Performance comparison of the raw feature sets on 10fold cross-validation using the... [/fig_ref] : first, among all the heterogeneous feature sets, the CLBP-Raw heterogeneous feature set served as the best performer in terms of all the evaluation metrics, suggesting the superiority of CLBP-Raw over the other four heterogonous feature sets. For example, CLBP-Raw achieved the F1-Score M ¼ 0.8084 and MCC ¼ 0.7857, which were 14.62% and 16.89%, 1.07% and 1.09%, 3.09% and 3.51%, and 4.22% and 4.75% higher than SLFs-Raw, LBP-Raw, RICLBP-Raw and LET-Raw, respectively. Upon closer inspection of the other four heterogeneous feature sets, we found that the LBP-Raw and RICLBP-Raw achieved the second and third best performance, respectively; Second, the performance of SLFs-Raw was not satisfactory. A possible reason is that the extracted raw heterogeneous feature sets might have redundant and noisy information which can result in the decreased predictive capabilities of the model; Third, keeping in mind that all the extracted raw heterogeneous feature sets may have redundant and noisy information, the performance results in [fig_ref] Table 2: Performance comparison of the raw feature sets on 10fold cross-validation using the... [/fig_ref] suggest that all the five heterogeneous feature sets examined in this study can be effectively used to predict protein subcellular localization.
## Performance comparison of optimal heterogeneous feature sets
As described in the Section 2.3, we fed each of the five extracted heterogeneous feature sets (i.e. SLFs-Raw, LBP-Raw, CLBP-Raw, RICLBP-Raw and LET-Raw) into the SDA FS algorithm and obtained its corresponding optimal heterogeneous feature set. Next, to investigate the discriminative capability of each optimal heterogeneous feature set, we performed 10-fold cross-validation on PScL2708 with SAE-SM as the classifier. provides the performance results of all the five selected optimal heterogeneous feature sets (i.e. SLFs-Optimal, LBP-Optimal, CLBP-Optimal, RICLBP-Optimal and LET-Optimal) in terms of OA, Rec M , Prec M , F1-Score M and MCC.
From , it can be observed that use of the SDA FS algorithm indeed helped improve the predictive performance of protein subcellular localization. Particularly, for each of the optimal heterogeneous feature sets, all the evaluation metrics (i.e. OA, Rec M , Prec M , F1-Score M and MCC) were improved compared with the raw heterogeneous feature sets. Moreover, among all the extracted raw and optimal heterogeneous feature sets, CLBP-Optimal [fig_ref] Figure 4: Performance comparison between the raw and optimal heterogeneous feature sets [/fig_ref] -C show the bar-graph representations of the OA, F1-Score M and MCC values of the raw and optimal heterogeneous feature sets. We can see that SAE-SM trained using the optimal heterogeneous feature sets consistently achieved a better performance than its counterpart trained using the raw heterogeneous feature sets. More specifically, from [fig_ref] Figure 4: Performance comparison between the raw and optimal heterogeneous feature sets [/fig_ref] , it can be seen that the predictive performance of the SAE-SM in term of OA on SLFs-Optimal, LBP-Optimal, CLBP-Optimal, RICLBP-Optimal and LET-Optimal was improved by 10.07%, 2.29%, 2.25%, 1.07% and 2.55%, respectively, in comparison to that of the SAE-SM achieved on SLFs-Raw, LBP-Raw, CLBP-Raw, RICLBP-Raw and LET-Raw. Similarly, as shown in [fig_ref] Figure 4: Performance comparison between the raw and optimal heterogeneous feature sets [/fig_ref] and C, the F1-Score M and MCC values were also improved based on the optimal heterogeneous feature sets.
In summary, the results in and [fig_ref] Figure 4: Performance comparison between the raw and optimal heterogeneous feature sets [/fig_ref] -C clearly demonstrate that utilizing the optimal heterogeneous feature sets consistently improved the predictive performance of our proposed SAE-SM method. Therefore, in the following sections, instead of utilizing the raw heterogeneous feature sets, we used SLFs-Optimal, LBP-Optimal, CLBP-Optimal, RICLBP-Optimal and LET-Optimal as our five optimal heterogeneous feature sets to construct our model.
## Integrating feature sets via 2l-sae-sm to improve the prediction performance
Supplementary Text S10 provides the performance comparison of different classifiers based on serial integration of the optimal heterogeneous feature sets. In this section, we seek to examine the effectiveness of the proposed 2L-SAE-SM method and investigate whether integrating all the optimal heterogeneous feature sets with 2L-SAE-SM can be even more effective than the direct serial integration for predicting protein subcellular localization. To address this, we performed experiments on the PScL2708 using 10-fold cross-validation and the performance comparisons between the best optimal heterogeneous feature set, the serially integrated feature set and the feature set integrated by 2L-SAE-SM are provided in [fig_ref] Table 4: Performance comparison between the best optimal features, features integrated serially and by... [/fig_ref].
From [fig_ref] Table 4: Performance comparison between the best optimal features, features integrated serially and by... [/fig_ref] , we can readily observe that the optimal heterogeneous feature sets integrated by 2L-SAE-SM drastically improved the performance. Compared with the feature set obtained by direct serial integration, the feature set integrated by 2L-SAE-SM achieved 3.07%, 3% and 3.62% improvements OA, F1-Score M and MCC, respectively. Similarly, by comparing with the CLBP-Optimal feature set, the feature set integrated by 2L-SAE-SM showed improvements of 6.17%, 6.61% and 7.28% in OA, F1-Score M and MCC, respectively. Additionally, the feature set integrated by 2L-SAE-SM also consistently performed well in terms of Rec M and Prec M .
In order to verify the efficacy of the proposed 2L-SAE-SM, we compared the performance of these three feature sets in terms of ROC and PR curves in Figures 5A-C and 6A-C. In particular, [fig_ref] Figure 5: ROC curves of the optimal heterogeneous features and features integrated serially and... [/fig_ref] , 5B and 6B and 5C and 6C show the ROC and PR curves for the CLBP-Optimal feature set, the serially integrated feature set and the feature set integrated by 2L-SAE-SM, respectively. Compared with the CLBP-Optimal and serially integrated feature sets, the feature set integrated by 2L-SAE-SM consistently achieved better performance by improving the ROC curve and AUC value for each individual class as shown in [fig_ref] Figure 5: ROC curves of the optimal heterogeneous features and features integrated serially and... [/fig_ref] -C. In terms of meanAUC, feature set integrated by 2L-SAE-SM achieved the meanAUC of 0.9906 and stdAUC of 0.0033 which was improved by 1.56% and 0.49% in comparison to the meanAUCs of 0.9750 and 0.9857 achieved by CLBP-Optimal and serially integrated feature sets, respectively. In addition, the feature set integrated by 2L-SAE-SM achieved the stdAUC of 0.0033 which was decreased by 0.77% and 0.15% in compassion to the stdAUCs of 0.0110 and 0.0048 achieved by CLBP-Optimal and serially integrated feature sets, respectively. Similarly, from [fig_ref] Figure 6: PR curves for optimal heterogeneous feature, features integrated serially and by 2L-SAE-SM [/fig_ref] -C, the feature set integrated by 2L-SAE-SM achieved improvements in terms PR curve of AUPR value for each individual class. Considering the meanAUPR and stdAUPR, the feature set integrated by 2L-SAE-SM achieved the meanAUPR and stdAUPR of 0.9608 and 0.0171, respectively, which is clearly better than the meanAUPR and stdAUPR of 0.8923 and 0.0518 achieved by CLBP-Optimal feature set and the meanAUC and stdAUPR of 0.9326 and 0.0273 achieved by serially integrated feature set. Altogether, the results and performance comparisons in terms of all evaluation indices in [fig_ref] Table 4: Performance comparison between the best optimal features, features integrated serially and by... [/fig_ref] , Figures 5A-C and 6A-C prove that integrating all the optimal feature sets by 2L-SAE-SM can indeed improve the prediction accuracy for protein subcellular localization.
Based upon the proposed 2L-SAE-SM, we developed a new computational method termed PScL-2LSAESM for the prediction of protein subcellular localization.
## Performance comparison with the other existing methods
In this section, to further illustrate the predictive power of PScL-2LSAESM, we performed experiments and compared its . Performance comparison of the optimal feature sets on 10-fold cross-validation using the benchmark training dataset PScL2708 performance with that of the other existing protein subcellular localization predictors including PScL-DDCFPred [bib_ref] PScL-DDCFPred: an ensemble deep learning-based approach for characterizing multiclass subcellular localization of..., Ullah [/bib_ref] , PScL-HDeep [bib_ref] PScL-HDeep: image-based prediction of protein subcellular location in human tissue using ensemble..., Ullah [/bib_ref] , SAE-RF [bib_ref] Bioimage-based prediction of protein subcellular location in human tissue with ensemble features..., Liu [/bib_ref] , SC-PSorter [bib_ref] Human cell structure-driven model construction for predicting protein subcellular location from biological..., Shao [/bib_ref] as well as the method proposed by [bib_ref] Image-based classification of protein subcellular location patterns in human reproductive tissue by..., Yang [/bib_ref].
## Performance comparison on 10-fold cross-validation test
In this section, we further compared the proposed PScL-2LSAESM with the other existing predictors by conducting 10-fold crossvalidation test on the PScL2708 dataset. To show the predictive capability, we first compared PScL-2LSAESM with the most recently published PScL-DDCFPred predictor. [fig_ref] Table 5: Performance comparison between PScL-2LSAESM and PScL-DDCFPred on 10-fold cross-validation using the benchmark... [/fig_ref] and -D show the experimental results of the two predictors. From [fig_ref] Table 5: Performance comparison between PScL-2LSAESM and PScL-DDCFPred on 10-fold cross-validation using the benchmark... [/fig_ref] and -D, we conclude that the proposed PScL-2LSAESM outperformed the recently published PScL-DDCFPred. In particular, PScL-2LSAESM achieved the OA, F1-Score M and MCC of 90.25%, 0.8953 and 0.8851, respectively, which were improved by 1.85%, 1.14% and 2.24% compared with the OA, F1-Score M and MCC of the PScL-DDCFPred predictor, respectively. In addition, the Rec M and Prec M of the PScL-2LSAESM were consistently improved as well.
Further, upon close inspection of and C which show the ROC and PR curves of PScL-DDCFPred and and D which show the ROC and PR curves of PScL-2LSAESM, respectively, we can see that the performance of the proposed PScL-2LSAESM was also clearly improved in terms of both ROC and PR curves. As shown in , the AUC values achieved by PScL-2LSAESM were consistently improved across all the subcellular localization classes compared with those of PScL-DDCFPred shown in . Similarly, upon a closer look at the results in , we can see that the proposed PScL-2LSAESM also achieved better AUPR curve values for the majority of the subcellular localization classes in comparison to the results in for PScL-DDCFPred. In addition, the meanAUC and meanAUPR values of the recently published PScL-DDCFPred were 0.9821 and 0.9417, respectively, while the meanAUC and meanAUPR values of the proposed PScL-2LSAESM were 0.9906 and 0.9608, respectively, the latter of which was clearly improved. Moroever, PScL-DDCFPred attained the stdAUC and stdAUPR values of 0.0037 and 0.0144, while the proposed PScL-2LSAESM method attained the stdAUC and stdAUPR of 0.0033 and 0.0171. Although the stdAUPR value of the proposed PScL-2LSAESM was not improved compared with PScL-DDCFPred, it achieved a competitive performance. In summary, the performance improvements in terms of all the other evaluation indices including OA, Rec M , Prec M , F1-Socre M , MCC, ROC curves and its AUC, meanAUC and stdAUC values, PR curves and its AUPR, meanAUPR values suggest the superiority of the proposed PScL-2LSAESM method over the previously developed PScL-DDCFPred method.
Next, we benchmarked the predictive capability of the proposed PScL-2LSAESM against that of PScL-HDeep, SAE-RF, SC-PSorter and Yang et al.'s method in [fig_ref] Table 6: Performance comparison of PScL-2LSAESM and the other existing methods on 10-fold cross-validation... [/fig_ref] in terms of OA, meanAUC and stdAUC. We can see that PScL-2LSAESM consistently achieved a better performance with an improvement of about 4.3-12.63% and 0.88-2.45% in OA and meanAUC, respectively, than the OA and meanAUC values of the other existing predictors. Similarly, the stdAUC of the PScL-2LSAESM was also the lowest.
In conclusion, the benchmarking results on the PScL2708 dataset confirmed that the proposed PScL-2LSAESM method achieved the best performance for the prediction of single-label multiclass protein subcellular localization.
## Performance comparison on the independent test
In this section, we performed independent test to further assess the generalization capability of the proposed PScL-2LSAESM. For this purpose, we first trained PScL-2LSAESM on the PScL2708 dataset and then tested the performance of the trained PScL-2LSAESM model on the independent PScL227 dataset. Next, we compared the [fig_ref] Table 7: Performance comparison between PScL-2LSAESM and PScL-DDCFPred on the independent test dataset PScL227Data... [/fig_ref]. From [fig_ref] Table 7: Performance comparison between PScL-2LSAESM and PScL-DDCFPred on the independent test dataset PScL227Data... [/fig_ref] , we can see that the proposed PScL-2LSAESM method achieved an improved performance than PScL-DDCFPred in terms of all the evaluation indices. For example, PScL-2LSAESM achieved the OA, F1-Score M and MCC of 74.88%, 0.7471 and 0.7085, respectively, which were 2.6%, 2.01% and 3.91% higher than the OA, F1-Score M and MCC values of PScL-DDCFPred, respectively. These results highlight that PScL-2LSAESM outperformed PScL-DDCFPred and provided a better generalization capability.
To further illustrate the generalization capability of PScL-2LSAESM, we provide the performance comparison of the proposed PScL-2LSAESM, PScL-HDeep, SAE-RF, SC-PSorter and Yang et al.'s method in terms of OA in . Note that parts of the results in were excerpted from the previous works of [bib_ref] Bioimage-based prediction of protein subcellular location in human tissue with ensemble features..., Liu [/bib_ref] and [bib_ref] PScL-HDeep: image-based prediction of protein subcellular location in human tissue using ensemble..., Ullah [/bib_ref]. We can see that PScL-2LSAESM attained the OA of 74.88% which was 11.9%, 8.76%, 7.73% and 3.86% higher than that of Yang et al.'s method, SC-PSorter, SAE-RF and PScL-HDeep, respectively. From , we conclude that PScL-2LSAESM has the best generalization capability compared with the other compared methods.
# Discussion
In this study, we have proposed a new computational method termed PScL-2LSAESM to effectively improve the bioimage-based prediction of human protein subcellular localization in human tissues. Specifically, a 2L-SAE-SM system was developed to integrate multiple heterogeneous feature sets into PScL-2LSAESM. For this purpose, we first designed a stacked based autoencoder with the SoftMax as the classifier layer, referred to as SAE-SM. Using SAE-SM, we implemented a two-level SAE-SM (2L-SAE-SM) where in the first level, each of the optimal heterogeneous feature sets (i.e. SLFs-Optimal, LBP-Optimal, CLBP-Optimal, LET-Optimal and RICLBP-Optimal) was fed into a single SAE-SM and then output the decision level set termed the 'intermediate decision' set. All the produced intermediate decision sets were then integrated using the ME method in the 2L-SAE-SM as the 'intermediate feature' set and then sent to the secondlevel SAE-SM. Both stringent 10-fold cross-validation test on the newly collected benchmark training dataset PScL2708 and independent test on the newly collected independent test dataset PScL227 have demonstrated the effectiveness of the proposed 2L-SAE-SM method for heterogeneous feature set integration. Extensive benchmarking experiments have also shown that the proposed PScL-2LSAESM predictor clearly outperformed the other existing single-label multiclass protein subcellular localization prediction methods. We expect that the proposed predictor can be explored as a useful method to facilitate the characterization of single-label multiclass protein subcellular localizations. In the future work, we plan to develop improved strategies to improve the performance of the proposed framework through the integration of multiple data sources such as protein amino acid sequences, protein images and protein-protein interaction networks.
[fig] Figure 1: Illustration of the SAE-SM-model architecture Text S6 provides a detailed description of the training procedures of 2L-SAE-SM. [/fig]
[fig] Figure 2: The architecture of the proposed 2L-SAE-SM model for integrating heterogeneous feature sets [/fig]
[fig] Figure 3: Illustration of the architecture of the proposed PScL-2LSAESM model consistently achieved the maximal performance of OA, Rec M , Prec M , F1-Score M and MCC, which were 84.08%, 82.83%, 83.02%, 0.8292 and 0.8123, respectively. For a fair comparison, [/fig]
[fig] Figure 4: Performance comparison between the raw and optimal heterogeneous feature sets. Panels (A-C) display the performance comparisons in terms of OA, F1-ScoreM and MCC, respectively [/fig]
[fig] Figure 5: ROC curves of the optimal heterogeneous features and features integrated serially and by 2L-SAE-SM. Panel (A) shows the ROC curves for the CLBP-Optimal feature set, panel (B) shows the ROC curves for the serially integrated feature set and panel (C) shows the ROC curves [/fig]
[fig] Figure 6: PR curves for optimal heterogeneous feature, features integrated serially and by 2L-SAE-SM. Panel (A) shows the PR curves for the CLBP-Optimal feature set, panel (B) shows the PR curves for the serially integrated feature set and panel (C) shows the PR curves for the feature set integrated by 2L-SAE-SM [/fig]
[fig] Figure 8, Figure 7: Performance comparison between PScL-2LSAESM and the existing predictors on the independent test dataset Performance comparisons between PScL-DDCFPred and the proposed PScL-2LSAESM method. Panels (A) and (B) show the ROC Curves, while (C) and (D) show the PR curves of PScL-DDCFPred and PScL-2LSAESM, respectively [/fig]
[table] Table 1: Statistical distribution of the images across each subcellular location class [/table]
[table] Table 2: Performance comparison of the raw feature sets on 10fold cross-validation using the benchmark training dataset PScL2708 [/table]
[table] Table 5: Performance comparison between PScL-2LSAESM and PScL-DDCFPred on 10-fold cross-validation using the benchmark training dataset PScL2708Results were calculated using the source code and data of PScL-DDCFPred. [/table]
[table] Table 4: Performance comparison between the best optimal features, features integrated serially and by 2L-SAE-SM on 10-fold cross-validation using the benchmark training dataset PScL2708 performance of PScL-2LSAESM with that of the recently published PScL-DDCFPred in terms of OA, Rec M , Prec M , F1-Score M and MCC. The performance results are provided in [/table]
[table] Table 6: Performance comparison of PScL-2LSAESM and the other existing methods on 10-fold cross-validation using the benchmark training dataset PScL2708Data excerpted fromLiu et al. (2020).Data excerpted fromUllah et al. (2021). [/table]
[table] Table 7: Performance comparison between PScL-2LSAESM and PScL-DDCFPred on the independent test dataset PScL227Data excerpted fromUllah et al. (2022). [/table]
|
Upper Limb Posture Estimation in Robotic and Virtual Reality-Based Rehabilitation
New motor rehabilitation therapies include virtual reality (VR) and robotic technologies. In limb rehabilitation, limb posture is required to (1) provide a limb realistic representation in VR games and (2) assess the patient improvement. When exoskeleton devices are used in the therapy, the measurements of their joint angles cannot be directly used to represent the posture of the patient limb, since the human and exoskeleton kinematic models differ. In response to this shortcoming, we propose a method to estimate the posture of the human limb attached to the exoskeleton. We use the exoskeleton joint angles measurements and the constraints of the exoskeleton on the limb to estimate the human limb joints angles. This paper presents (a) the mathematical formulation and solution to the problem, (b) the implementation of the proposed solution on a commercial exoskeleton system for the upper limb rehabilitation, (c) its integration into a rehabilitation VR game platform, and (d) the quantitative assessment of the method during elbow and wrist analytic training. Results show that this method properly estimates the limb posture to (i) animate avatars that represent the patient in VR games and (ii) obtain kinematic data for the patient assessment during elbow and wrist analytic rehabilitation.
# Introduction
Robotic and VR technologies are important components of the modern neurorehabilitation systems for pathologies such as stroke or spinal cord injury [bib_ref] Motor learning: its relevance to stroke recovery and neurorehabilitation, Krakauer [/bib_ref] [bib_ref] A virtual realitybased exercise system for hand rehabilitation post-stroke, Adamevich [/bib_ref] [bib_ref] Robot-aided neurorehabilitation of the upper extremities, Riener [/bib_ref]. In this field, our general research has two main goals:
(a) to improve the assessment of the rehabilitation progress through precise estimation of the patient kinematics. This is the focus of this paper;
(b) to optimize the rehabilitation processes by using the kinematic (and other) patient models. This optimization includes hybrid technologies (e.g., robotics, virtual reality, functional electrical stimulation [bib_ref] Advanced hybrid technology for neurorehabilitation: the HYPER project, Mauro [/bib_ref] , etc.). Even though this domain is very important for rehabilitation, we see it as a natural consequence of (a) and we concentrate on (a) at this time.
In the mentioned scenario, the proper estimation of the patient limb posture is a fundamental prerequisite for the following:
(1) design and control of the advanced robotic exoskeletons which provide assistance to the patient during motor rehabilitation [bib_ref] On the kinematic design of exoskeletons and their fixations with a human..., Jarrassé [/bib_ref] [bib_ref] Admittance control of an upper limb exoskeleton-reduction of energy exchange, Kim [/bib_ref] ,
(2) animation of realistic avatars representing the patient in virtual reality (VR) scenarios (e.g., games, bionics), and
(3) acquisition of kinematic data of the patient during the training exercises to assess improvement along the therapy.
This paper presents a method for estimation of limb posture from the exoskeleton posture. Notice that such an estimation is not trivial, since the limb is not rigid, is not standard, and has kinematic topology different from the exoskeleton topology.
Our method delivers limb postures estimates to strengthen and to enable downstream applications in robotic rehabilitation (among others, using VR [bib_ref] Advanced hybrid technology for neurorehabilitation: the HYPER project, Mauro [/bib_ref].
## Robotic-based motor rehabilitation
Therapy. The inclusion of robotic devices in motor rehabilitation therapies has been increasing over the last decade. The robot-assisted therapies complement conventional rehabilitation by providing intensive, repetitive, task-specific, and interactive treatment. All these factors contribute to a more effective rehabilitation [bib_ref] A robotic system to train activities of daily living in a virtual..., Guidali [/bib_ref] [bib_ref] Positive effects of robotic exoskeleton training of upper limb reaching movements after..., Frisoli [/bib_ref] [bib_ref] A robotic device as a sensitive quantitative tool to assess upper limb..., Gilliaux [/bib_ref].
Robotic-assisted therapy has been shown to improve active movement, strengthening, and coordination in stroke patients [bib_ref] Effects of robotic therapy on motor impairment and recovery in chronic stroke, Fasoli [/bib_ref]. The majority of clinical studies have reported that robot-assisted therapy can ease impairments and lower disabilities of the affected patient [bib_ref] Poststroke upper extremity rehabilitation: a review of robotic systems and clinical results, Brewer [/bib_ref]. Moreover, evidence suggests that task-oriented exercises using robotic devices produce significant improvements in recovering lost abilities [bib_ref] Whats new in stroke rehabilitation, Teasell [/bib_ref].
Combining these exercises with VR games makes the therapy more attractive to the patient, increasing motivation and treatment effects [bib_ref] Advanced hybrid technology for neurorehabilitation: the HYPER project, Mauro [/bib_ref] [bib_ref] Virtual arm representation an d multimodal monitoring for the upper limb robot..., Epelde [/bib_ref]. It is important that these games are designed to be consistent with the principles of physical therapy and adjustable to the level of impairment [bib_ref] Exoskeleton robots for upper-limb rehabilitation: state of the art and future prospects, Lo [/bib_ref].
A central element in designing a therapy is the feedback that patients receive. To achieve relatively permanent changes in the capability of producing skilled action, it is crucial to provide the patient with proper feedback in order to produce a positive impact on the neural mechanism promoting motor learning.
Feedback includes all the sensory information as the result of a movement and it is divided into two classes: (1) intrinsic or inherent feedback, which is information captured by human sensory systems as a result of the normal production of the movement, and (2) extrinsic or augmented feedback, which is information that supplements intrinsic feedback. Robotic-assisted therapy with VR games including animated realistic avatars may improve the quality and specificity of extrinsic feedback that the patient receives.
From the perspective of the therapist, robotic devices can be used to obtain quantitative metrics for the assessment of the improvement of the patient. The kinematic information of the affected limb during the exercises is required to compute several evaluation metrics, such as joint amplitudes, speeds, movement smoothness, and directional control.
## Case study armeo exoskeleton.
Our proposed therapy uses the Armeo Spring exoskeleton for the upper limb intervention [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref]. We find the following limitations of this system.
(1) Currently, the gaming platform provides an elementary assessment of the patient performance with metrics such as Hand Path Ratio [bib_ref] A kinematic analysis of a haptic handheld stylus in a virtual environment:..., Broeren [/bib_ref] and joint range of motion, which are only available in certain games of the Armeo proprietary platform. We propose a continuous quantification of the patient performance along the treatment therapies, involving metrics that are highly correlated with the functional recovery of the patient.
(2) Currently, the games only provide the patient with feedback of his hand position. We propose to provide a 3-dimensional representation of the arm, which would help the patient to immerse in the VR environment.
The kinematic data provided by the exoskeleton samples the angular position of its joints. Such information cannot be used directly to represent the human arm, since the patient limb and the exoskeleton kinematic models differ significantly.
This paper presents a method to estimate the posture of the limb by using the kinematic data provided by the exoskeleton. We propose to solve the limb's inverse kinematics (IK) problem extended with the kinematic constraints of the exoskeleton fixations on the limb. This extended problem is solved in real time with standard robotic libraries. In this manner, we aim to overcome the limitations of the Armeo system regarding to the feedback and assessment of the patient.
This paper is organized as follows: Section 2 presents a brief literature review. Section 3 addresses the formal statement of the problem and the proposed method to solve it. Section 4 discusses the implementation of our approach and its use in VR games. Section 5 presents the evaluation methodology of our approach in the realm of motor rehabilitation. Section 6 informs and discusses the results of the experiments conducted using our solution strategy. Section 7 concludes the paper and identifies future developments.
## Literature review
Several estimation methods and human models have been proposed in the literature to solve the problem of limb posture estimation. Next, we present a brief review of developments in these areas. We define a free movement scenario as a situation in which the patient limb does not wear an exoskeleton or interact with any other robotic interface. Under the mentioned conditions, the literature that addresses upper limb posture estimation considers tasks in which the human subject has to reach a desired object. Therefore, these approaches are designed to estimate the posture of the upper limb based on a given target position and orientation of the hand.
Statistical [bib_ref] Human motion simulation for vehicle and workplace design, Chaffin [/bib_ref] [bib_ref] Statistics for digital human motion modeling in ergonomics, Faraway [/bib_ref] , IK [bib_ref] An inverse kinematics architecture enforcing an arbitrary number of strict priority levels, Baerlocher [/bib_ref] [bib_ref] Redundancy resolution of the human arm and an upper limb exoskeleton, Kim [/bib_ref] [bib_ref] Two-handed human reach prediction models for ergonomic evaluation, Jung [/bib_ref] , and direct optimization [bib_ref] 3D human lifting motion prediction with different performance measures, Xiang [/bib_ref] [bib_ref] Multi-objective optimisation method for posture prediction and analysis with consideration of fatigue..., Ma [/bib_ref] [bib_ref] Optimization-based posture prediction for human upper body, Mi [/bib_ref] [bib_ref] Synthesizing redundancy resolution criteria of the human arm posture in reaching movements, Li [/bib_ref] [bib_ref] Realtime optimal reach-posture prediction in a new interactive virtual environment, Yang [/bib_ref] [bib_ref] Posture reconstruction-some insights, Gragg [/bib_ref] methods are the most used approaches to estimate the limb posture [bib_ref] Validation of predicted posture for the virtual human santos, Yang [/bib_ref].
Statistical or data-based approaches model the human kinematics with regressive models from empirical data [bib_ref] Multi-objective optimization for upper body posture prediction, Yang [/bib_ref]. Factors such as the size of the database of captured motions [bib_ref] A comparison between optimization-based human motion prediction methods: data-based, knowledge-based and hybrid..., Pasciuto [/bib_ref] and the characteristics of the population involved in the experiments impact the accuracy and usefulness of these models.
Kinematic approaches model the human limbs with links, joints of different degrees of freedom, and end-effectors [bib_ref] Realtime optimal reach-posture prediction in a new interactive virtual environment, Yang [/bib_ref]. The IK problem is then solved with either closed-form or numerical methods. The quality of the kinematic model and the convergence speed and robustness of the approach used to solve the IK problem directly affect the accuracy of the estimations.
Optimization approaches require a nontrivial function to minimize, which actually leads to the desired configuration (typically, a minimal energy one [bib_ref] A comparison between optimization-based human motion prediction methods: data-based, knowledge-based and hybrid..., Pasciuto [/bib_ref]. When optimization is used to solve an IK problem, additional constraints can be easily included in the formulation [bib_ref] Synthesizing redundancy resolution criteria of the human arm posture in reaching movements, Li [/bib_ref] [bib_ref] Realtime optimal reach-posture prediction in a new interactive virtual environment, Yang [/bib_ref] [bib_ref] Posture reconstruction-some insights, Gragg [/bib_ref].
Approaches combining optimization-based and statistical models have been also proposed to overcome the individual limitations of optimization and statistical methods [bib_ref] A comparison between optimization-based human motion prediction methods: data-based, knowledge-based and hybrid..., Pasciuto [/bib_ref] [bib_ref] Hybrid predictive dynamics: a new approach to simulate human motion, Xiang [/bib_ref]. Naturally, the composed method requires a highquality dataset of motions and the formulation of proper objective and constraints functions.
## Robotic-assisted scenario.
There is a shortage in the literature addressing posture estimation of the human limb while interacting with an exoskeleton. Although exoskeletons are designed with the ultimate goal of minimizing their kinematic differences with human limbs and interact seamlessly with them, the following factors influence the human motion patterns and therefore the posture of the limb:
(1) the mechanic design of the exoskeleton (inertia, back drivability, friction, joint motion limits, etc.).
(2) the type of assistance that the exoskeleton provides (passive, active, and assist-when-needed).
(3) the performance of the exoskeleton motion controller.
Here, using a naive one-to-one mapping between the joint angles of the human limb and exoskeleton leads to poor positioning results [bib_ref] A method for measuring the upper limb motion and com puting a..., Jarrasse [/bib_ref].
References [bib_ref] Admittance control of an upper limb exoskeleton-reduction of energy exchange, Kim [/bib_ref] [bib_ref] Redundancy resolution of the human arm and an upper limb exoskeleton, Kim [/bib_ref] propose the computation of the arm's IK by using a disambiguation criteria for its redundancy which chooses a swivel angle such that the palm points to the head region. This methodology is suitable for real-time implementation and it is used in the control strategy of the active 7-DOF exoskeleton developed by the authors' research team [bib_ref] Upper-limb powered exoskeleton design, Perry [/bib_ref]. The authors report that the mean error in the estimation of the swivel angle is less than 5 degrees. The magnitude of the errors in the estimation of the wrist, elbow, and GH-joint angles is not reported.
References [bib_ref] Admittance control of an upper limb exoskeleton-reduction of energy exchange, Kim [/bib_ref] [bib_ref] Redundancy resolution of the human arm and an upper limb exoskeleton, Kim [/bib_ref] do not consider the motions of the clavicle and scapula (which affect the position of the GHjoint center) in the estimation of the posture of the arm, as they assume the position of the GH-joint center to be known. Therefore, this approach should not be used in cases in which the position of the GH-joint center cannot be determined from data provided by the exoskeleton (e.g., Armeo Spring) or by any additional motion capture system.
Other common methods to estimate the posture of human limbs cannot be used or are impractical in roboticassisted scenarios. For example, inertial and magnetic measurement systems (IMMSs) presented in [bib_ref] Ambulatory measurement of shoulder and elbow kinematics through inertial and magnetic sensors, Cutti [/bib_ref] [bib_ref] Reducing drifts in the inertial measurements of wrist and elbow positions, Zhou [/bib_ref] are unusable because the magnetic disturbances produced by the metallic components of the exoskeleton corrupt the magnetic sensor measurements.
If optical tracking systems are used, arrays of markers need to be attached to the patient in order to measure the limb joint angles. Occlusions of such markers are frequently produced by the mechanic structure of the exoskeleton when performing the rehabilitation exercises. To overcome the occlusions of the markers, a redundant setup is necessary [bib_ref] Validation of predicted posture for the virtual human santos, Yang [/bib_ref]. This limitation makes the use of optical tracking systems cumbersome for frequent use in the rehabilitation therapy.
## Human model.
A central element in human posture estimation is the human kinematic model itself. Simple models based on hierarchies of links and lower kinematic pairs can be found in [bib_ref] Realtime optimal reach-posture prediction in a new interactive virtual environment, Yang [/bib_ref] [bib_ref] Reach envelope of a 9-degree-of-freedom model of the upper extremity, Yang [/bib_ref] [bib_ref] Determining weights of joint displacement objective function for standing reach tasks, Zou [/bib_ref] [bib_ref] An alternative formulation for determining weights of joint displacement objective function in..., Zou [/bib_ref] [bib_ref] Hybrid method for driver accommodation using optimization-based digital human models, Gragg [/bib_ref]. These approaches results are convenient for real-time tasks and for implementation. However, more elaborated models should be used to describe complex kinematic relationships [bib_ref] Qualitative and quantitative descriptions of glenohumeral motion, Hill [/bib_ref] , such as the shoulder rhythm [bib_ref] Human shoulder modeling including scapulo-thoracic constraint and joint sinus cones, Maurel [/bib_ref]. On the other hand, musculoskeletal models reported in [bib_ref] A model of the upper extremity for simulating musculoskeletal surgery and analyzing..., Holzbaur [/bib_ref] [bib_ref] Multibody biomechanical models of the upper limb, Ambrósioa [/bib_ref] [bib_ref] Musculoskeletal upper limb modeling with muscle activation for flexible body simulation, Kim [/bib_ref] offer better accuracy for dynamics computations, since they include forces from muscles and ligaments.
The selection of the human kinematics model rests not only on the kinematic statement of the problem, but also on the compromise between accuracy and speed required in a particular application.
## Conclusions of literature review.
Although the methods designed to estimate the posture of the upper limb (in absence of a robotic interface) reviewed in Section 2.1.1 could be used in robotic-assisted rehabilitation, we have not found any actual implementation of them in this context. Usage of these methods without any change in their design parameters in robotic-assisted applications may lead to erroneous posture estimations, given the influence of the exoskeleton on human motion patterns. Therefore, the validity of these methods in the robotic-assisted scenario remains to be proven. An additional limitation of these methods is that only few of them have been validated quantitatively by determining the errors in their estimations.
On the other hand, the few posture estimation approaches that address limb interaction with an exoskeleton (Section 2.1.2) have been designed to specifically solve the arm posture estimation problem, limiting their usability in posture estimation of other human limbs.
In response to the mentioned issues, in this paper we present the following:
(1) a method that can be applied, in a general manner, to solve the limb posture estimation problem using kinematic data provided by the exoskeleton attached to the limb,
# Materials and methods
## Problem description.
In this section, we state the problem of estimating the joint angles of the patient limb during robotic-assisted rehabilitation therapy from the kinematic information provided by the robot. The elements that are considered inputs to the problem are the following: (1) the geometry and topology (e.g., the Denavit-Hartenberg parameters [bib_ref] A kinematic notation for lower-pair mechanisms based on matrices, Denavit [/bib_ref] of the exoskeleton and the human limb, (2) a known configuration of the angles of the joints of the exoskeleton, (3) the kinematic constraints imposed by the fixations of the exoskeleton over the patient limb (which result from wearing the exoskeleton), and (4) the constraints that govern the posture of the patient limb while interacting with the exoskeleton, which are related to mechanical and control factors of the exoskeleton that influence the patient movement. The goal of the proposed algorithm is to find the approximate joint angles of the patient limb, such that the mentioned constraints are met. This problem can be formally stated as follows.
## Given
(1) the kinematic model of the exoskeleton ( , ), where and are sets of links and joints, respectively, (d) the vector ∈ R , = ∑ =0 ( ), contains the set of independent coordinates that defines a configuration of uniquely:
[formula] (i) = {V 0 , . . . , V , . . . , V }; (ii) [/formula]
represents the state of in instant and its value is known;
(2) a human patient with a kinematic model of his limb ( , ), where and are sets of links and joints, respectively, (d) the vector ∈ R , = ∑ =0 ( ), contains the set of independent coordinates that defines a configuration of uniquely:
[formula] (a) = { 0 , . . . , +1 }; (b) = { 0 , . . . , }: (i) ( ) denotes the DOF of ; (ii) V = { 1 , . . . , ( ) } is(i) = {V 0 , . . . , V , . . . , V }; (ii) the th element of , , is subject to ℎ ( ) = min ≤ ≤ max ( ∈ [0, − 1]); (iii) [/formula]
represents the state of in instant and its real value is unknown;
## Goal is as follows
(1) to find the vector̃∈ R , which approximates such that
[formula] (a) ( , ) = 0 ∀ ∈ [0, ]; (b) ℎ ( ) = min ≤ ≤ max ∀ ∈ [0, − 1]; (c) ( ) = 0 ∀ ∈ [0, ]. [/formula]
To solve this problem, a method based on IK of the limb has been developed. The following sections describe the methodology implemented.
## Kinematic
Modeling of the Exoskeleton. The Armeo Spring [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] is a passive exoskeleton (orthosis) that supports the weight of the arm of the patient. The level of support provided by the system springs can be adjusted, regulating the effort of the patient arm to overcome gravity. The exoskeleton has a total of seven angle sensors to measure the position of its rotational joints and one pressure sensor to measure the gripping force at the hand.
We built a kinematic model of the Armeo Spring [fig_ref] 2: the implementation of our proposed method for the upper limb posture estimation... [/fig_ref] , which contains both prismatic and revolute joints. The prismatic joints of the exoskeleton allow adjusting it to the different sizes of the patients, and they remain fixed during the training.
Our implementation models the links and joints of the Armeo exoskeleton and creates a hierarchical structure of them.
Although the Armeo exoskeleton presents a parallelogram mechanism in its kinematic chain, the exoskeleton can be modeled with a serial chain extended with a dependency equation among the joints used to represent the parallel mechanism. [fig_ref] Figure 3: Human upper body kinematic model [/fig_ref] shows the kinematic model of the human upper body that we created for this application. The joints of the model are represented with green color. The upper limb is highlighted using links in light green color. Our upper body model (33-DOF) includes joints of the spine, shoulder complex, elbow, and wrist. It is based on the ones presented in , which have been widely used in the area of human posture estimation. The main advantages of those models are their easy implementation and their suitability for solving the posture estimation problem in real time, which is one of the main requirements of our application. A weakness of those kinematic models is that the glenohumeral (GH) joint is modeled with a kinematic chain of three concurrent revolute joints, orthogonal to each other. In this way, the rotation of the GH joint is parameterized with Euler angles and suffers from gimbal lock [bib_ref] Practical parameterization of rotations using the exponential map, Grassia [/bib_ref]. In order to avoid this limitation, the GH joint is represented in our model with a spherical joint, such that other rotation parameterizations (e.g., quaternion or exponential map) can be used.
## Kinematic modeling of the human upper body.
Although there are more complex and accurate kinematic models of the upper body, the results obtained in [bib_ref] An alternative formulation for determining weights of joint displacement objective function in..., Zou [/bib_ref] , in a scenario where the subject does not interact with an exoskeleton in an application that is not related to motor rehabilitation, show that posture estimations for the upper limb can be obtained with a reasonable accuracy by using their original model.
The neutral or rest posture of the arm is defined with the arm fully extended along the body as in [bib_ref] The human arm kinematics and dynamics during daily activities-toward a 7 DOF..., Rosen [/bib_ref]. The range of motion of the joints of the arm obtained in [bib_ref] Upper-limb powered exoskeleton design, Perry [/bib_ref] (derived from a motion study during the execution of activities of daily living) is used as reference to establish the joint limits of our model, which correspond to constraint 2(d)(ii) in the list presented in Section 3.1. There are several factors that affect the satisfaction of the constraints during the execution of the exercises. This set includes (1) deformation of the coupling mechanisms and (2) uncertainty or errors in the modeling of the human upper limb. Therefore, these constraints are exactly met only under ideal conditions and in practice they do not capture all the details of the real interaction. However, as we prove, they suffice to obtain a reasonable accuracy in the estimation of the limb posture.
## Modeling the kinematic constraints of
## Arm constraint.
The arm fixation imposes a position (3-DOF) constraint on the human arm. The point on the arm that follows the position of the fixation is determined by an initialization process between the and kinematic chains (see Section 3.6).
In our model, the fixations are modeled as rigid bodies. However, the exoskeleton fixations are made of flexible materials, such that their geometry is deformed when large forces are applied on them.
The arm fixation suffers significant deformation when the arm is moved towards a horizontal configuration (e.g., when performing a complete stretching of the arm along the sagittal or frontal plane). In those cases, the coordinate system at the exoskeleton arm fixation center undergoes a translation, resulting from the deformation of the fixation mechanism that is not reproduced by our model.
To deal with this kind of situations, the weights of constraints representing fixations that suffer less deformation than other ones are adjusted such that they receive more importance when solving the IK problem. In this way, the limb posture is estimated meeting the constraints that model with more fidelity the observed behavior. In this case, the weight of the arm constraint is lower than the ones belonging to the forearm and arm restrictions. [fig_ref] Figure 4: Interaction of the Upper Limb and the Exoskeleton [/fig_ref] shows the human arm (blue transparent cylinder) with the fixation of the exoskeleton for the arm (black transparent ring) around it. The constraint imposed by this fixation to the arm is represented by the matching of (a) human arm (white disk) versus (b) fixation (yellow disk) coordinate systems. Figures 4(a) and 4(b) correspond to unsatisfied and satisfied constraints, respectively.
## Forearm constraint.
The forearm fixation imposes a 3-DOF position constraint on the human forearm. The point on the human forearm that moves together with the fixation is determined in the initializing stage. Additionally, the fixation is able to rotate around its longitudinal axis, according to the forearm pronation/supination movement (1-DOF orientation constraint). The rotation angle is measured with an encoder. The forearm constraint forces the human wrist flexion/extension axis to be approximately aligned with the exoskeleton's wrist joint axis.
## Hand constraint.
The hand constraint forces the human hand to follow the position and orientation (6-DOF) of the Armeo hand grip. The patient exercises while grabbing the handle of the exoskeleton. The mechanic design of the Armeo avoids the slippage of the hand with respect to the axis of the handle during the execution of the exercises. As with the previous fixations, the point on the hand where the coordinate system of the hand is located is calculated in the initialization stage. Mismatch of the human and robot coordinate systems : State of the kinematic chains before the initialization process (notation in .
[formula] A B b a C D F c [/formula]
Currently, it is implemented to attract the GH joint to a position (3-DOF position constraint) below the first joint of the Armeo ( 0 joint represented with symbol in , which does not suffer any translation during the training of the patient. By keeping the GH joint near 0 comfortable postures for the spine and arm can be achieved. [fig_ref] Figure 5 3: [/fig_ref] shows that the shoulder constraint prevents the excessive motion of the joints of the spine, which is a compensatory movement that should be also avoided during the rehabilitation therapy. The shoulder constraint is central to proper posture estimation during shoulder abduction.
## Inverse kinematics. given a desired pose (position and orientation) vector
∈ R 6 for the end-effector of an open kinematic chain , the IK problem is to find the vector of angles of the robot's joints ∈ R (where corresponds to the DOFs of ), such that the difference = − between and the actual pose of the end-effector of , ∈ R 6 , approaches zero.
There are several approaches to solve this problem, including analytic [bib_ref] Closed-form solution for inverse kinematics of robot manipulators with redundancy, Chang [/bib_ref] and numerical methods [bib_ref] Selectively damped least squares for inverse kinematics, Buss [/bib_ref]. The iterative strategy used to solve the IK problem is based on the Jacobian matrix of the manipulator ( ), which linearly relates the velocity of the end-effector and the joints bẏ = ( )̇.
(1) : Glossary related to the .
Symbol Description 0 2 Arm fixation coordinate system Forearm fixation coordinate system Armeo hand grip coordinate system Human arm end-effector coordinate system Human forearm end-effector coordinate system Human hand end-effector coordinate system By replacing Δ for in [bib_ref] A virtual realitybased exercise system for hand rehabilitation post-stroke, Adamevich [/bib_ref] , which is obtained by discretizing (1), the necessary Δ to approximate is obtained:
[formula] Δ = ( ) −1 Δ .(2) [/formula]
Notice that ( ) may not be square (consider, e.g., a kinematic chain with more than 6-DOF) or invertible. In those cases, the pseudoinverse and damped least squares (DLS) methods (among others) can be used to obtain Δ , such that ‖ ‖ is minimized. The pseudoinverse method is computationally faster than the DLS but tends to be unstable when the robot approaches a singular configuration. The DLS method offers more robustness (specially when is out of reach) at the cost of a slower convergence.
## Relation among end-effectors and targets.
The aforementioned strategy to solve the IK problem can also be used in situations in which the manipulator has more than one end-effector. In this case, the error vector is given by = { 1 − 1 , . . . , − , . . . , Nee − Nee } where Nee is the number of end-effectors of the robot. Notice that vector = − is not necessarily a point ∈ R 6 . For example, if only the position (and not the orientation) of the th end-effector is specified, ∈ R 3 .
In our application, the formulation of the IK problem with multiple end-effectors and targets can be used to represent the constraints discussed in Section 3.4. In this way, each constraint can be represented by a target and end-effector pair. The coordinate frames of the end-effectors ( ) ( ∈ [1, . . . , Nee]) are attached to the human limb, so their position and orientation depend on the current configuration of the limb, . The coordinate frames of the targets of the limb ( ) [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref]. . , Nee]) are attached to the exoskeleton such that they are transformed according to its current configuration . Then, the IK problem is solved for the limb, finding such that = ‖ ( )− ( )‖ ≈ 0 ( ∈ [1, . . . , Nee]). Notice that if represents a kinematic constraint, ∈ R ( ) where ∈ [0, ]. Otherwise, represents a restriction related to the natural posture of the limb, and therefore ∈ R dim( ) , where ∈ [0, ], and Nee = + + 2. Notice that, due to modeling inaccuracies of the kinematic chains or the constraints, it is possible that for a configuration some constraints cannot be satisfied within the desired tolerance. That situation can be interpreted as if some targets ( ) are not reachable. It is important that the method used to solve the IK problem handles this situation robustly, avoiding oscillations. For this reason the DLS method was used.
## Joints and constraint weights.
References [bib_ref] Determining weights of joint displacement objective function for standing reach tasks, Zou [/bib_ref] [bib_ref] An alternative formulation for determining weights of joint displacement objective function in..., Zou [/bib_ref] state that giving more importance to some of the model joints over others, by assigning weights to the joints, allows estimating more accurately the posture of the human limb.
Let us assume that is the weight of joint and that joints and can contribute to the movement of endeffector to diminish . Then, if > , the displacement that performs is larger than the one done by . This means that is preferred to be moved over to reach a desired target.
In our model, the weights of the joints of the upper body were adjusted such that the joints on the spine of the model perform small displacements in comparison with the movement performed by the shoulder, elbow, and wrist joints.
On the other hand, applying weights to the error vector gives more importance to reach a specific target over others. In our approach, this translates into giving some constraints more importance than others. Let us define with ( ∈ [0, ]) the weight of the constraint and with ( ∈ [0, ]) the weight of constraint.
In our model, high weights were adjusted for the kinematic constraints imposed by the exoskeleton fixations ( ≈ 1.0). Otherwise, low weights ( ≈ 0.2) were assigned to the other type of constraints.
There are different formulations of the DLS method that incorporate weights for the joints and error vector (e.g., [bib_ref] Robust inverse kinematics using damped least squares with dynamic weighting, Schinstock [/bib_ref]. In V-REP, the following DLS formulation is used to solve . This also applies for weights related to constraints . However, independent weights can be assigned for the position and orientation components of a constraint.
## Initialization of the kinematic chains.
To accurately estimate the limb posture, it is required to properly couple the human and exoskeleton kinematic models. To do so, we require to correctly position the end-effectors of the human kinematic model with respect to the arm, forearm, and hand coordinate systems. These end-effectors must be positioned such that they are able to move together with the coordinate systems of the fixations of the exoskeleton model (targets). Notice that the position of the end-effectors with respect to the links of the human model changes according to the actual patient and exoskeleton dimensions. depicts a state in which the human and exoskeleton models are decoupled. The correct position and orientation of the coordinate systems of the end-effectors of the human model have not been calculated, and, therefore, they do not match the position and orientation of the exoskeleton's fixations coordinate systems.
The initialization of the kinematic chains requires a reference pose of the exoskeleton in which (a) the human joints angles can be determined accurately and (b) the exoskeleton's fixations undergo negligible deformation, reducing the uncertainty about the position of the human model endeffectors.
The pose of the exoskeleton that meets the mentioned requirements is the one in which the flexion/extension of the shoulder and elbow take place in the sagittal plane . In this pose, the position of the human GH joint with respect to the exoskeleton base can be easily determined because the joints of the spine and shoulder complex are in their rest position.
The coupling process involves the following steps.
(1) Position the exoskeleton model such that the joint 0 lies above the human GH joint. Adjust the height of the exoskeleton model such that 2 is at the level of the human GH joint. These instructions are prescribed by the manufacturer of exoskeleton to use it with the actual patient.
(2) Compute the arm flexion and abduction angles such that the arm passes through the origin of the arm fixation coordinate system. Adjust the origin of the arm end-effector coordinate system to match the origin of the arm fixation.
(3) With the position of the elbow joint defined, compute the elbow flexion and the GH internal rotation angles such that the forearm passes through the origin of the exoskeleton forearm fixation. Adjust the origin of the forearm end-effector coordinate system to match the origin of the forearm fixation. (4) Compute the wrist extension angle such that the human hand is able to grasp the exoskeleton's hand grip. Adjust the hand end-effector to match the position of the Armeo's end-effector at the hand grip.
(5) Calculate the forearm pronation/supination angle such that the wrist's extension/flexion axis matches the orientation of the Armeo's hand grip longitudinal axis with respect to the human forearm pronation/supination axis.
(6) Adjust the human forearm and hand end-effector coordinate systems to match the orientation of the forearm and Armeo's end-effector coordinate systems, respectively.
The result of the initialization process is depicted in [fig_ref] Figure 7: Result of the initialization procedure [/fig_ref].
## Implementation
To implement the proposed method the virtual robot experimentation platform (V-REP) was used, which is an open source robotics simulator. V-REP provides tools to easily and efficiently create kinematic models of rigid multibody systems and to solve IK problems. Using the simulator, a scene was created, which contains both the human upper body and Armeo kinematic models [fig_ref] 2: the implementation of our proposed method for the upper limb posture estimation... [/fig_ref]. The weights of the human kinematic model were adjusted (Section 3.5.2) and the simulator's IK module was configured to include the kinematic constraints (Section 3.4).
The source code of the simulator was compiled, modified, and integrated into our rehabilitation platform. Custom classes and functions were programmed to allow easy data exchange among the Armeo, the rehabilitation game platform, and the IK module of the simulator.
The limb posture estimation process consists of the following steps.
(1) Obtain the angles of the Armeo's joints by using hardware and software interfaces provided by HOCOMA AG.
(2) Use the obtained angles to update the joints angles of the Armeo's kinematic model in the simulator.
(3) Retrieve the angles of the joints of the human model computed by the simulator's IK module.
Computing the inverse kinematics of our upper limb kinematic model, once the Armeo model is updated in the simulator with the real joint measurements of the exoskeleton, takes less than 4 ms on a 2.13 Ghz dual-core CPU. Therefore, the implemented method is suitable for realtime posture estimation without using high-performance hardware.
After the joint estimates are produced, we use them to update the patient avatar in VR games. We also store them in a database for a posterior patient assessment. presents a user test of the limb posture estimation algorithm feeding the Armeo kinematic model in the simulator (in real time) with the Armeo Spring joint positions measured by its encoders. This figure presents the posture of the test subject and Armeo Spring in parallel with estimations of the user posture in the simulator. The test subject performed (a) reaching exercises, in which the subject recreated the postures of his arm to reach and grab objects that are close to his body . These exercises are frequently practiced during the arm rehabilitation;
(b) extreme region exercises, in which the subject positioned his hand in the boundaries of his arm workspace . These exercises are challenging for the subject and are less likely to occur during the therapies due to the exercises difficulty.
## Vr games.
Currently, we have implemented two types of games for the robotic-assisted upper limb rehabilitation therapy. The first type of games focuses on the rehabilitation of reaching movements. The second type of games addresses the rehabilitation of analytic movements of the GH, elbow, and wrist joints.
## Reaching rehabilitation.
Reaching rehabilitation is performed by training the movements that are required to reach and grasp objects with the hand. These exercises involve several joints of the upper limb, and, therefore, they are considered complex.
To train these exercises, we have programmed a game in which the patient controls the movement of a virtual human arm by moving his own arm . The target of the patient is to reach a specific object (e.g., cube) in the scene, grab it, and bring it to a releasing area (e.g., green circle).
## Analytic movements rehabilitation.
According to motor learning theories, the training of analytic movements constitutes the first step into learning complex motor tasks. In such a step, simple movements involving few DOFs of the limb are practiced [bib_ref] Freezing degrees of freedom in skill acquisition, Vereijken [/bib_ref].
For this scenario, we have programmed a game in which the patient controls the position of a spaceship, along the horizontal axis of the screen, by performing 1-DOF movements with the wrist, elbow, or GH joint. The target of the game is that the patient positions the spaceship under an alien that moves along a vertical path from the top to the bottom of the screen. When the position of the spaceship is correct, it fires a gun and destroys the alien. For both games, the limitations of the mobility of the patient are identified in a calibration phase, guarantying that the target of the games is properly located. Other game parameters (number of executions, max execution time per task, target size, etc.) are adjusted through the medical interface . The medical interface allows the physician to select the games for the training, configure its parameters, and review metrics related to the performance of the patient during a game.
The VR games were programed with the OpenScene-Graph API, which allows animating deformable virtual objects and creating scenes with dynamic simulations using the Bullet Physics package. The graphic rendering of the VR game runs at 30 frames per second using a ATI Radeon HD 4600 GPU, which is a midrange graphic card.
During the therapy, the patient sees the VR scene. The kinematic models are used for IK computations and they are not displayed.
## Evaluation
In order to determine the accuracy of our developed method, the joint angles of 4 voluntary healthy male test subjects (average age 34 years) were measured by using an optical tracking system and compared with the angles obtained from our posture estimation algorithm during the execution of typical (in this case, analytic movements) robotic-assisted rehabilitation exercises. As discussed in Section 4.1.2, the rehabilitation of analytic movements is a necessary step before addressing the rehabilitation of complex motor tasks.
The specific exercises performed by the test subjects were The evaluation of our method has been conducted without performing any previous setting or automatic adjustment of the weights or other parameters of the approach in order to reduce the estimation errors. However, algorithm training might be used in the future to improve the method's performance.
## Measurement of the upper limb joint angles.
A detailed explanation of the method that was used to measure the human joint angles would merit an additional manuscript. Nevertheless, a basic description of this method is provided next.
In order to measure the limb joint angles of the test subject, we use a Polaris Spectra optical tracking system (OTS). In order to track the limb movements, it is necessary to install on test subject limb a set of rigid bodies with passive markers. By detecting these passive markers (reflective spheres), the OTS is able to compute the position and orientation of each rigid body. One rigid body (reference rigid body) is used as the coordinate system of reference for the measurements of the OTS. The position and orientation of the other rigid bodies (mobile rigid bodies) are computed with respect to the reference rigid body.
The reference and mobile rigid bodies are installed on different arm segments (i.e., upper arm, forearm, and hand) according to the joint angle to be measured. [fig_ref] Table 2: Installation of the reference and mobile rigid bodies in the evaluation [/fig_ref] shows the installation of the reference and mobile rigid bodies for each of the joint angles that we measured. [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] shows the configuration of the rigid bodies to measure the elbow flexion/extension angle.
In order to measure the human joint angles, we have adapted the method presented in [bib_ref] Functionally interpretable local coordinate systems for the upper extremity using inertial &..., De Vries [/bib_ref] , which is originally proposed to be used with IMMSs, to implement it by using an OTS. In [bib_ref] Functionally interpretable local coordinate systems for the upper extremity using inertial &..., De Vries [/bib_ref] it is proposed to measure the joint angles by following the next steps.
(1) Compute a reference coordinate system for the joint of interest. A subset of the axes of the resulting coordinate system match the axes of rotation of the joint. The position and orientation of the joint coordinate system are defined with respect to the reference rigid body.
(2) Compute the orientation of the mobile rigid body with respect to the joint coordinate system.
(3) Compute the joint angles that result from rotations of the mobile rigid body by using Euler-angles decomposition. The rotations of the mobile rigid body are caused by the exercising of the subject joint.
To build an orthogonal right-handed coordinate system of reference for the joint, we identify each axis of rotation of the joint, as proposed in [bib_ref] Functionally interpretable local coordinate systems for the upper extremity using inertial &..., De Vries [/bib_ref].
To identify each rotation axis of the joint, we use the instant helical axis method described in [bib_ref] Orientation of axes in the elbow and forearm for biomechanical modelling, Veeger [/bib_ref]. A rotational axis of the joint is computed from the kinematic data of the mobile rigid body while the subject performs 1-DOF movements of the joint.
In contrast to the proposal presented in [bib_ref] Functionally interpretable local coordinate systems for the upper extremity using inertial &..., De Vries [/bib_ref] to compute the wrist joint coordinate system, we build this coordinate system by identifying only the flexion/extension axis, given that the ulnar/radial deviation cannot be trained with the Armeo Spring.
Accuracy of the Limb Joint Angles Measurement Method. In motor rehabilitation, goniometry is the standard method to measure the angle at the patient joints. This is a manual method, and, therefore, its efficacy depends on the examiner experience [bib_ref] Analysis of the reliability and reproducibility of goniometry compared to hand photogrammetry, De Carvalho [/bib_ref]. One of the limitations of this method is that it provides a resolution (minimal detectable change) in measuring limb joint angles of about 8 degrees [bib_ref] The reliability and concurrent validity of scapular plane shoulder elevation measurements using..., Kolber [/bib_ref]. In other words, this method should not be used to measure angles smaller than 8 degrees because in those cases measurements present large uncertainty.
Alternative approaches to measure the patient limb joint angles are IMMS-based methods. One of the methods that provide better accuracy than goniometry is presented in [bib_ref] Ambulatory measurement of shoulder and elbow kinematics through inertial and magnetic sensors, Cutti [/bib_ref]. This method provides a measurement accuracy characterized by a RMSE of less than 3.6 degrees. The authors of the mentioned work conclude that this accuracy is proper for measuring elbow and shoulder angles of clinical relevance in ambulatory settings.
In tests with an artificial 1-DOF joint, the method to measure the limb joint angles that we have adapted from [bib_ref] Functionally interpretable local coordinate systems for the upper extremity using inertial &..., De Vries [/bib_ref] allowed us to estimate the joint angle with a RMSE smaller than 1 degree. According to a comparison with the accuracy provided by the reviewed methods, we conclude that the method proposed by [bib_ref] Functionally interpretable local coordinate systems for the upper extremity using inertial &..., De Vries [/bib_ref] to measure the limb posture is valid to determine the accuracy of our proposed limb posture estimation method. [fig_ref] Table 3: Summary of main features of the evaluation tests [/fig_ref] summarizes the main features of the evaluation that we have conducted.
## Protocol.
For each trial of the evaluation exercises we performed the following steps.
(1) Compute the joint coordinate system corresponding to the evaluation exercise (Section 5.1). (2) Instruct the subject to perform the corresponding evaluation exercise until the number of desired joint angle measurements is taken. During the execution of the evaluation exercises the amplitude, speed and the number of cycles of the movements in each trial were left to the discretion of each test subject. In the evaluation, the VR games were not executed, given that they are not necessary to assess the accuracy of the posture estimation algorithm. Furthermore, in this way the influence of the VR games on the subject movement amplitude, speed, and repetitions is avoided, which derives a richer variety of movement features in the evaluation exercises. However, it is worth mentioning that the joint limits of the exoskeleton, the need to avoid occlusions of the passive markers on the rigid bodies attached to the test subject, and the limited detection volume of the OTS do constrain the subject's movements.
# Results and discussion
In this section, we present the results of the experiments described in Section 5. Tables 4, 6, 7, and 8 (angles expressed in degrees) present the average RMSE obtained in the estimation of the angle of interest by using our proposed algorithm. Each table presents the average ROM of the movement performed by each test subject. The average RMSE and ROM metrics mentioned previously are obtained from the 4 trials that each subject performed for each exercise. The last row in the tables presents the average values of each of the computed metrics for all subjects.
N.B.: in this section we compare our results against freely moving subject cases reported in the literature. We resort to such free movement cases since we found no reports concerning estimations errors of the wrist or elbow angles in limbs constrained with exoskeletons. presents angle estimation statistics for wrist flexion and extension. The ROM exercised by the subjects presents small variability and seems not to correlate with RSME. However, we did observe that subjects 1 and 2 performed slow movements while subjects 3 and 4 moved fast. Such a difference reflects on the RMSE values.
To elaborate this point, we present in [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] the history of the measured versus estimated angle, for subjects 1 and 3. The sampling span is 250 (approx. 3.75 seconds). The motion features of the movements shown in [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] are summarized in [fig_ref] Table 5: Motion features for subjects 1 and 3 in WFE exercise [/fig_ref]. In such table, the time delay aspect refers to the time delay that the estimations provided by our algorithm present with respect to the measured angles. The time delay is larger when the subject moves fast. This causes the increment in the RMSE estimation values.
These results suggest that the response speed of our algorithm, given a change in the Armeo joint angles caused by the movement of the human subject, allows providing better estimates when the subject moves slowly (as in rehabilitation therapy). In our algorithm, the response speed largely depends on the damping constant used in the DLS method to solve the limb's IK. By using a smaller damping constant in the DLS method, the response speed can be improved, sacrificing some stability.
Nevertheless, the average RMSE obtained for all subjects shows a better performance of our method with respect to [bib_ref] An alternative formulation for determining weights of joint displacement objective function in..., Zou [/bib_ref] , an optimization-based approach which presents errors around 3.5 degrees. Compared to [bib_ref] Reducing drifts in the inertial measurements of wrist and elbow positions, Zhou [/bib_ref] , which presents a IMMS-based method to estimate the wrist angles with a RMSE of less than 3 degrees, our results are slightly better.
## Elbow flexion/extension.
In flexion and extension of elbow [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] , [fig_ref] Table 6: Estimation errors in elbow flexion/extension exercise [/fig_ref] , involuntary movement along the pronation/supination axis is not avoided. Therefore, small excursions in this DOF were observed.
For all subjects, our method overestimates the amplitude of rotational movements about the flexion/extension axis, when compared against the measured values (see [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] for subject 2).
Our method performs better than the one in [bib_ref] An alternative formulation for determining weights of joint displacement objective function in..., Zou [/bib_ref] , in which the reported mean error in estimating the flexion/extension angle is approximately 14 degrees. Compared to the approach in [bib_ref] Ambulatory measurement of shoulder and elbow kinematics through inertial and magnetic sensors, Cutti [/bib_ref] , which uses a IMMS-based method and presents a RMSE of 3.6 degrees in estimating elbow and shoulder angles, our method also presents better performance.
We include in [fig_ref] Table 6: Estimation errors in elbow flexion/extension exercise [/fig_ref] the estimation statistics for pronation/supination angle in order to illustrate the performance of our method with small angular displacements. [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] displays the estimation and measurement of pronation/supination angle for a trial of subject 2. In this figure, we observe that there is an underestimation of the angle. However, it must be taken into account that estimation errors for small ROMs are in the same order of the measurement method accuracy (RMSE 1 degree). [fig_ref] Table 7: Estimation errors in forearm pronation/supination exercise [/fig_ref] and [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] show the statistics of our method for forearm pronation/supination angle estimation. We remark that motion in the elbow flexion/extension axis may occur during the forearm pronation/supination exercise. Therefore, we also report (in [fig_ref] Table 7: Estimation errors in forearm pronation/supination exercise [/fig_ref] and [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] the estimation results for the small angular movements around the flexion/extension axis. The average RMSE in the estimation of the pronation/supination DOF of our method presents an accuracy similar to the one of [bib_ref] Ambulatory measurement of shoulder and elbow kinematics through inertial and magnetic sensors, Cutti [/bib_ref] (RMSE 3.6 degrees). [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] shows the elbow angles estimation results for a trial of the FPS exercise of subject 1. [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] shows that estimations in the flexion/extension DOF, in which small movements were performed, do not present the oscillations of the measured angle (RMSE 1.175 degrees). On the other hand, [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] shows that estimations of the pronation/supination angle are very close to the measured values.
## Forearm pronation/supination.
For the pronation/supination angle, the worse estimations were obtained for subject 4, who performed short but very fast movements, affecting the estimation accuracy as described in Section 6.1. According to results presented here and in Section 6.2, it seems that for small movements the estimation approach is slightly more sensitive to movements in the pronation/supination DOF than on the flexion/extension DOF.
## Simultaneous elbow flexion/extension and forearm
Pronation/Supination. The objective of this exercise is to evaluate how simultaneous movements of both DOFs of the elbow affect the angle estimations for this joint. The results are presented in [fig_ref] Table 8: Estimation errors in simultaneous elbow flexion/extension and forearm pronation/supination exercise [/fig_ref]. In this table, it is shown that, for both elbow DOFs, the average of the RMSE for all the subjects is similar to the one presented in [bib_ref] Ambulatory measurement of shoulder and elbow kinematics through inertial and magnetic sensors, Cutti [/bib_ref] (RMSE 3.6 degrees).
This result also suggests that, during the performance of a functional rehabilitation exercise, such as reaching, in which simultaneous flexion/extension and pronation/supination movement are necessary, the accuracy of the estimations would remain in an adequate range. [fig_ref] Figure 1: Armeo Spring orthosis [/fig_ref] presents the estimation results of a trial of this exercise of subject 4. In this figure, it can be observed that estimations follow closely the measured angles.
## Conclusions and future work
This paper presents a method that can be applied to estimate the posture of the human limbs during the interaction with exoskeletons by solving the limb IK problem extended with the kinematic constraints of the exoskeleton fixations on the limb. The few approaches in the literature that deal with limb posture estimation in a robotic-assisted scenario are specifically designed to estimate the arm posture. In contrast, the method that we propose provides a general formulation, which is not specific to any human limb or exoskeleton. Our method is based on inverse kinematics and it can be implemented using standard robotics libraries.
In this paper, we have also shown the implementation of the method to provide upper limb posture estimations, in real time, using the Armeo Spring. We have also presented the use of the resulting limb postures estimations in the animation of avatars in VR rehabilitation games.
We have evaluated the accuracy of the estimations of our method during the performance of analytic rehabilitation exercises of the wrist and elbow. The obtained results show that our approach presents an accuracy that is better than the one provided by goniometry, which is the traditional method to measure the patient angles in motor rehabilitation. Compared to the accuracy provided by IMMSs-based methods, which are considered enough accurate to measure clinical relevant limb joint angles in nonrobotic-assisted scenarios, we have obtained very similar results.
Based on the mentioned results, we conclude that our approach can be used to (a) provide an estimation of the pose of the human upper limb with enough accuracy to be used for avatar animation in VR games and (b) obtain the kinematic data for the patient assessment during analytic training of the elbow and wrist.
Future work includes (a) the exploration of other approaches to model the flexible fixations of the exoskeleton, (b) the definition of a set of weights for the human model joints that represent the movement features of a set of human subjects, and (c) a quantitative assessment of the performance of our method in a functional rehabilitation scenario.
## Nomenclature
Clavicle: One of the bones of the shoulder girdle. It is located at the root of the neck DLS:
Damped least squares DOF:
Degree of freedom EFE:
Elbow flexion/extension FPS:
Forearm pronation/supination GH: Glenohumeral Humerus: Upper arm bone IK:
Inverse kinematics IMMSs: Inertial and magnetic measurement systems OTS:
Optical tracking system RMSE:
Root mean square error ROM:
Range of motion Scapula: One of the bones of the shoulder girdle. It connects the humerus with the clavicle SEFEFPS: Simultaneous EFE and FPS VR: Virtual reality V-REP: Virtual robot experimentation platform WFE:
Wrist flexion/extension V:
Total number of constraints of the IK problem (V ∈ N) :
IK error vector ( ∈ R V ) :
T o t a lD O F so ft h eh u m a nk i n e m a t i cm o d e l ( ∈ N) :
Total DOFs of the exoskeleton kinematic model ( ∈ N) :
Jacobian matrix of the IK problem ( ∈ R V× ) :
V × V identity matrix :
Diagonalma trixofjoin tsweigh ts ( ∈ R + × ) :
Diagonal matrix of constraints weights ( ∈ R + V×V ) :
Vector of joint angles of the human kinematic model in instant ( ∈ R ) :
Vector of joint angles of the exoskeleton kinematic model in instant ( ∈ R ) :
Damping factor of DLS method ( ∈ R + ).
[fig] Figure 1: Armeo Spring orthosis. [/fig]
[fig] 2: the implementation of our proposed method for the upper limb posture estimation using the Armeo Spring exoskeleton, and (3) the quantitative validation of our proposed method by determining the estimation errors during the training of meaningful upper limb rehabilitation exercises. [/fig]
[fig] Figure 3: Human upper body kinematic model. [/fig]
[fig] Figure 4: Interaction of the Upper Limb and the Exoskeleton. The Armeo provides fixations for the human limb. These fixations introduce constraints on the position and orientation of the coordinate systems attached to the arm, forearm, and hand. Constraint modeling the interaction of the Armeo's arm fixation. [/fig]
[fig] Figure 5: 3.4.4. Shoulder Constraint.The shoulder constraint does not belong to the set of movement restrictions imposed by the coupling mechanisms of the Armeo. Instead, it is related to the restrictions intended to produce a natural posture of the upper limb considering also the influence of the exoskeleton on the patient movements. This constraint helps to choose one of the multiple configurations of the human kinematic chain that comply with the other categories of constraints. Effect of the shoulder constraint in the upper limb posture estimation. [/fig]
[fig] Figure 7: Result of the initialization procedure. [/fig]
[fig] Figure 8, Figure 9: Test subject in parallel with estimations of his posture in the simulator. (a) shows reaching exercises and (b) shows extreme region exercises. Games and medical interface. [/fig]
[fig] 1: wrist flexion/extension (WFE), (2) elbow flexion/extension (EFE), (3) forearm pronation/supination (FPS), (4) simultaneous elbow flexion/extension and forearm pronation/supination (SEFEFPS). [/fig]
[fig] 3: Compute the RMSE in the estimation of each joint angle by comparing the measured angle with the estimation provided by our algorithm. (4) Compute the ROM of the subject movements from the measured angles. [/fig]
[table] Table 5: Motion features for subjects 1 and 3 in WFE exercise. [/table]
[table] Table 6: Estimation errors in elbow flexion/extension exercise (units in degrees). [/table]
[table] Table 7: Estimation errors in forearm pronation/supination exercise (units in degrees). [/table]
[table] Table 8: Estimation errors in simultaneous elbow flexion/extension and forearm pronation/supination exercise (units in degrees). [/table]
|
Talus Bipartitus: A Rare Anatomical Variant Presenting as an Entrapment Neuropathy of the Tibial Nerve within the Tarsal Tunnel
Tarsal tunnel syndrome is an entrapment neuropathy of the tibial nerve within the tarsal tunnel that lies beneath the retinaculum on the medial side of the ankle. It is often underdiagnosed. Talus bipartitus is a rare anatomical variant; only a few cases have been described in medical literature. We report a case of a 36-year-old female with tarsal tunnel syndrome secondary to a talus bipartitus undergoing surgical treatment with good clinical outcome. To our knowledge, talus bipartitus presenting as tarsal tunnel syndrome has no previous reports. Image studies and physical examination are crucial to reach precise diagnosis.
# Introduction
Tarsal tunnel syndrome (TTS) is an entrapment neuropathy of the tibial nerve or its branches within the tarsal tunnel that lies beneath the retinaculum on the medial side of the ankle. This condition is frequently underdiagnosed. Main symptoms are burning, cramping, and pain along the foot plantar region. The typical clinical profile includes worsening of symptoms as the day goes on and nocturnal awakening with tingling feet [bib_ref] The tarsal tunnel syndrome: a study of pressures within the tunnel and..., Nayagam [/bib_ref] [bib_ref] The four medial ankle tunnels: a critical review of perceptions of tarsal..., Dellon [/bib_ref].
Clinical tests include the dorsiflexion-eversion provocative manoeuvre in which the tibial nerve is compressed by positioning the ankle in passively maximally eversion and dorsiflexion while all of the metatarsophalangeal joints are maximally dorsiflexed and held in this position for five to ten seconds [bib_ref] The dorsiflexion-eversion test for diagnosis of tarsal tunnel syndrome, Kinoshita [/bib_ref]. Tinel's sign is present in most of the cases, having a 92% sensibility and 100% specificity, with a predictive value of 85%. The use of electromyography may also be helpful [bib_ref] The four medial ankle tunnels: a critical review of perceptions of tarsal..., Dellon [/bib_ref] [bib_ref] Tarsal tunnel syndrome: a literature review, Ahmad [/bib_ref]. TTS aetiology varies from heel varus, arthritis, tendinopathy, osteophytes, lipomas, perineural fibrosis, trauma, systemic diseases, and venous pathology to postsurgical injuries [bib_ref] Tarsal tunnel syndrome: a literature review, Ahmad [/bib_ref] [bib_ref] Results of surgical treatment of tarsal tunnel syndrome, Reichert [/bib_ref] [bib_ref] Outcome of surgical treatment of tarsal tunnel syndrome, Sammarco [/bib_ref]. There is also a considerable number of idiopathic cases.
Talus bipartitus (TB) is a rare anatomical anomaly of the talus. It consists of two talar fragments separated by the presence of articular cartilage. There are less than 30 reported cases of TB in medical literature. There is no information about the prevalence of this infrequent condition.
We present a case of a 36-year-old female with tarsal tunnel syndrome due to a talus bipartitus with a medial prominence, undergoing surgical treatment with good clinical outcome.
## Case report
A 36-year-old female patient presented with a 4-year history of right ankle pain. She had significate loss of weight and started physical activity just before the symptoms appeared four years ago. No history of trauma was reported. First diagnosis was synovitis, for which the patient underwent physical therapy with torpid outcome. She experienced paraesthesia along the posterior and medial aspect of the ankle and foot. Symptoms got worse at the end of the day, with cramping episodes in the foot. Physical examination revealed a positive dorsiflexion-eversion test provoking numbness of the foot. Tinel's sign was positive. Ankle range of motion was limited due to pain, scored 8 using the Visual Analog Score (VAS). The patient had a preoperative score of 40 points, according to the American Orthopaedic Foot and Ankle Society (AOFAS). Bilateral electromyography (EMG) was performed, disclosing abnormal adductor hallucis and adductor digiti quinti neurophysiologic parameters in comparison to the asymptomatic side.
Initial plain radiographs showed a considerable (1.8 cm) posterior bone fragment in relation with the talus . CT scan was then performed, which showed the presence of an articulated accessory bone. At this point, talus bipartitus (TB) diagnosis was suspected . MRI disclosed the presence of an inflammatory process, bone fragment covered with cartilage, and what appears to be a degenerative synchondrosis as a consequence of posteromedial impingement [fig_ref] Figure 3: MRI, sagittal view of the ankle [/fig_ref].
In this scenario, a surgical procedure was indicated due to increasing symptoms and several conservative treatment trials with no response.
Preoperatory assessment included three-dimensional CT scan images to get full awareness of the bone fragment's specific anatomy. Anatomic relationship between the tibial nerve, posterior tibial artery, flexor hallucis longus, flexor digitorium longus, and tibialis posterior tendon and the bone fragment was assessed with the help of MRI. Though fixation of the accessory bone has been previously reported, in the index case, this was not an option because of its particular shape, with a medial protuberance invading the tarsal tunnel.
Excision of the bone fragment was performed. A medial approach was used. Both tibialis posterior tendon and neurovascular bundle were identified and carefully preserved. The bone fragment was removed and the tunnel was released [fig_ref] Figure 4: Medial approach with exposure of the pseudoarticular space [/fig_ref]. Postoperative care included cast immobilization and a 30-day weight-bearing restriction. Postoperative radiographs showed the complete absence of residual bone fragment . Histological findings disclosed mature bone with articular cartilage surface. After 6 months of follow-up, the patient had a VAS score of 2 and an AOFAS score of 87. At a 2-year follow-up, VAS was 1 and AOFAS was 96. Patient had no residual pain and no recurrence of neurological symptoms.
# Discussion
All image studies suggested that the cause of our patient's pain and neurological condition was attributable to the presence of the unusual bone fragment. Furthermore, the particular anatomical shape of this bone fragment with a clear prominence towards the medial aspect of the ankle reinforced the diagnosis.
However, image studies were not able to confirm as a gold standard method whether this bone prominence was indeed entrapping any nerve structure or even in contact with. Though it may have seemed that diagnosis in this index case was easy to achieve, the patient lingered for a while with no diagnosis.
MRI is a good method for identifying pathologic causes of TTS, specific space occupying lesions [bib_ref] Tarsal tunnel syndrome: still more opinions than evidence. Status of the art, Doneddu [/bib_ref]. McSweeney and Cichero [bib_ref] Tarsal tunnel syndrome -a narrative literature review, Mcsweeney [/bib_ref] demonstrated in their study that MRI could identify the cause of TTS in 88% of patients with symptoms. That being said, TTS diagnosis is not always easy to achieve and is regularly underdiagnosed [bib_ref] Tarsal tunnel syndrome: a literature review, Ahmad [/bib_ref] [bib_ref] Tarsal tunnel syndrome: still more opinions than evidence. Status of the art, Doneddu [/bib_ref]. Physical examination is crucial. TTS specific cause can be identified in up to 80% of cases [bib_ref] Tarsal tunnel syndrome: diagnosis, surgical technique, and functional outcome, Bailie [/bib_ref] [bib_ref] Tarsal tunnel syndrome: review of the literature, Cimino [/bib_ref] ; provocative tests such as Tinel's sign are mandatory to reach neuropathy diagnosis [bib_ref] Provocative tests in different stages of carpal tunnel syndrome, Mondelli [/bib_ref] [bib_ref] Usefulness of electrodiagnostic techniques in the evaluation of suspected tarsal tunnel syndrome:..., Patel [/bib_ref]. When clinical findings are not conclusive, axonal injury can be demonstrated with an EMG. In this particular case, both MRI and EMG worked together to achieve proper diagnosis. EMG by itself will never spur a suspicion of any bone anatomical variation. MRI is helpful to study the anatomy of the tarsal tunnel and to reinforce TB diagnosis (presence of cartilage). CT scan is even a better method to reach TB diagnosis and to study its particular anatomy.
Reichert et al [bib_ref] Results of surgical treatment of tarsal tunnel syndrome, Reichert [/bib_ref]. reported a series of 31 patients with TTS undergoing surgical treatment. Only in 9 patients an internal cause was found (ganglion, neuroma, tendinopathy, and lipoma). They concluded that knowing the cause of TTS improves the effectiveness of treatment.
Surgical treatment of TTS is still challenging. Gondring et al. [bib_ref] An outcomes analysis of surgical treatment of tarsal tunnel syndrome, Gondring [/bib_ref] showed in their study that only 51% of the patients with TTS had better quality of life after surgical decompression. Other series have been reported, with only 42% satisfactory outcome in patients with TTS postsurgery [bib_ref] Tarsal tunnel syndrome: a study of the clinical and neurophysiological results of..., Ward [/bib_ref].
To our knowledge, only 1 case of TTS has been reported secondary to a skeletal anomaly (accessory ossicle) [bib_ref] Tarsal tunnel syndrome secondary to an unreported ossicle of the talus: a..., Sweed [/bib_ref]. TB is a very rare skeletal variation. Tsuruta et al. [bib_ref] Radiological study of the accessory skeletal elements in the foot and ankle..., Tsuruta [/bib_ref] performed a radiological study with more than 3000 feet without finding TB. According to medical literature, its aetiology still remains uncertain. It usually constitutes about one-third of the posterior aspect of the talar body and is separated from that structure by a frontal split [bib_ref] Talus bipartitus: a rare skeletal variation, Rammelt [/bib_ref].
In the clinical practice, TB is vaguely seen and most physicians tend to attribute symptoms to more frequent conditions. Differential diagnosis of TB includes the presence of an os trigonum, which lacks of articular surface and has a higher incidence (13%). Zwiers et al. [bib_ref] Talus bipartitus: a systematic review and report of two cases with arthroscopic..., Zwiers [/bib_ref] identified only 23 reported cases of TB in their systematic review. None of them involved neuropathy entrapment. Main symptoms described were pain, swelling, and restricted range of motion. Surgical treatment included removal or fixation of the accessory bone.
In this index case, surgical technique was focused in removing the bone fragment, preserving structures and recognizing any nerve injury if present. We cannot provide any reliable information about surgical approach to TTS release by reporting one single case, which may vary according to aetiology.
We present what it seems to be the first reported case of tarsal tunnel syndrome secondary to talus bipartitus. TTS is relatively common but has many different causes. Clinical findings and proper imaging are crucial to reach precise diagnosis and to conduct preoperatory planning.
# Ethical approval
All investigations were conducted in conformity with ethical principles of research.
[fig] Figure 1, Figure 2: X-ray of the hind foot, lateral view, showing a 1.8 cm accessory posterior bone fragment (arrow). Comparative CT scan showing posterior bone process, displacing into the medial side (arrow) in both coronal (a) and axial (b) view. [/fig]
[fig] Figure 4: Medial approach with exposure of the pseudoarticular space (star), tibialis posterior retracted forward (yellow arrow), and the accessory bone fragment (white arrow). [/fig]
[fig] Figure 3: MRI, sagittal view of the ankle; the presence of cartilage between the talar body and the accessory bone can be appreciated (stars). [/fig]
|
Plasmodium falciparum clearance in clinical studies of artesunate-amodiaquine and comparator treatments in sub-Saharan Africa, 1999–2009
Background: Artemisinin-based combination therapy (ACT) is the recommended first-line therapy for uncomplicated Plasmodium falciparum malaria worldwide but decreased artemisinin susceptibility, phenotypically characterized as slow parasite clearance time (PCT), has now been reported in Southeast Asia. This makes it all too important to measure the dynamics of parasite clearance in African patients treated with ACT over time, to understand trends and detect changes early enough to intervene Methods: Individual patient data from 27 clinical trials of artesunate-amodiaquine (ASAQ) vs comparators conducted between 1999 and 2009 were analysed for parasite clearance on modified intent-to-treat (ITT) basis.
Results: Overall 15,017 patients treated for uncomplicated P. falciparum malaria at 44 sites in 20 sub-Saharan African countries were included in the analysis; 51% (n=7,660) vs 49% (n=7,357) were treated with ASAQ and comparator treatments, respectively. Seventy-seven per cent (77%) were children under six years of age. The proportion of the patients treated with ASAQ with persistent parasitaemia on Day 2 was 8.6%, and 1.5% on Day 3. Risk factor for not clearing parasites on Day 2 and Day 3 calculated by multivariate logistic regression with random effect on site and controlling for treatment were: high parasitaemia before treatment was (adjusted risk ratios (AOR) 2.12, 95% CI 1.91-2.35, AOR 2.43, 95% CI 1.98-3.00, respectively); non-ACT treatment (p=0.001, for all comparisons). Anaemia (p=0.001) was an additional factor for Day 2 and young age (p=0.005) for Day 3. In patients treated with ASAQ in studies who had complete parasitaemia data every 24 hours up to Day 3 and additionally Day 7, the parasite reduction ratio was 93.9% by Day 1 and 99.9% by Day 2. Using the median parasitaemia before treatment (p0=27,125 μL) and a fitted model, the predicted PCT (pPCT = 3.614*ln (p0) -6.135, r² = 0.94) in ASAQ recipients was 31 hours.
Conclusion: Within the period covered by these studies, rapid Plasmodium falciparum clearance continues to be achieved in Sub-Saharan African patients treated with ACT, and in particular with ASAQ. The prediction formula for parasite clearance time could be a pragmatic tool for studies with binary outcomes and once-daily sampling, both for research and monitoring purposes.
# Background
Delayed Plasmodium falciparum response to artemisinin is reported in Southeast Asia, and may spread to other, more intensely malaria-endemic areas. Artemisinin compounds contribute rapid parasite killing and faster clearance to artemisinin-containing combination therapy (ACT). Artemisinin tolerance/resistance manifests itself with slower parasite clearance, while ACT remain generally effective both clinically and parasitologically [bib_ref] Artemisinin-resistant Plasmodium falciparum in Pursat province, western Cambodia: a parasite clearance rate..., Amaratunga [/bib_ref] [bib_ref] Reduced artemisinin susceptibility of Plasmodium falciparum ring stages in western Cambodia, Witkowski [/bib_ref] [bib_ref] Nosten F: Changes in the treatment responses to artesunate-mefloquine on the northwestern..., Carrara [/bib_ref]. Genomic studies are underway, which have generated so far two sets of putative markers of artemisinin resistance -K13-propeller polymorphism [bib_ref] A molecular marker of artemisininresistant Plasmodium falciparum malaria, Ariey [/bib_ref] , and a SNP in the gene encoding a DNA repair protein RAD5 on chromosome 13 (and possibly also on chromosome 10) [bib_ref] Novel phenotypic assays for the detection of artemisinin-resistant Plasmodium falciparum malaria in..., Witkowski [/bib_ref]. An in vitro assay (Ring Stage Assay) has also focused on the very early phases of the ring-stage parasite [bib_ref] Genetic loci associated with delayed clearance of Plasmodium falciparum following artemisinin treatment..., Takala-Harrison [/bib_ref]. However, as of today, in the clinic as well as in the field the parasite clearance time (PCT) and its related clinical phenotype (delayed PCT) remain the best practical surrogates of artemisinin in vivo resistance [bib_ref] Early parasitological response following artemisinin-containing regimens: a critical review of the literature, Das [/bib_ref] ; the problem is that the frequent sampling (every six or eight hours) required to measure PCT accurately and to estimate the parasite clearance half-life [bib_ref] Standardizing the measurement of parasite clearance in falciparum malaria: the parasite clearance..., Flegg [/bib_ref] is practically difficult even in research settings. Easier methods are needed for both research and routine purposes. For example, failure to clear by Day 3 (72 hours post-treatment start) is proposed as a simple and accurate predictor of treatment failure [bib_ref] In vivo parasitological measures of artemisinin susceptibility, Stepniewska [/bib_ref] , requiring a single time-point and thus limiting the workload in resource-limited settings.
Applying these and other measures both retrospectively and prospectively to commonly-used ACT will help understand trends and detect changes early enough to prompt effective responses to contain the spread of resistance. Artesunate-amodiaquine (ASAQ) is the second most commonly used ACT in the world (the first-line treatment in 22 out of 44 sub-Saharan African countries); more than 200 million treatments of ASAQ Winthrop have been distributed in Africa since the medication became available in 2007. While ASAQ is generally effective, the decreasing efficacy of AQ single-agent treatment have been reported in some sub-Saharan African settings [bib_ref] Efficacy of artesunate-amodiaquine for treating uncomplicated falciparum malaria in sub-Saharan Africa: a..., Zwang [/bib_ref] [bib_ref] Artemisinin-based combination therapy for treating uncomplicated malaria, Sinclair [/bib_ref]. Should artemisinin resistance occur there, it is expected to result soon in treatment failures, particularly in subjects with partial immunity, such as children in high-transmission areas, or patients of all ages in lowtransmission areas. Monitoring ASAQ efficacy, particularly in these groups, has therefore become all the more important.
Greater understanding of factors influencing parasite clearance is crucial, and requires the analysis of pooled data from individual patient records. Therefore, this study collected and analysed a large sample (over 15,000 patients) of individual patient data (of whom approximately three-quarters were children under six years of age) from clinical trials conducted in sub-Saharan Africa between 1999 and 2009. Studies such as this will help provide the foundations for further analyses of parasite clearance trends across sub-Saharan malaria-endemic countries.
# Methods
## Study sites, design and patients
The database was constructed from studies identified through a systematic review of clinical trials and personal contacts. To be included, efficacy or tolerability monitoring studies had to have been conducted in sub-Saharan Africa including any formulation of ASAQ to any other single or combination treatment of uncomplicated falciparum malaria (non-severe or non hyperparasitaemic) with follow-up of at least 28 days. For the studies meeting these criteria, investigators were contacted to provide individual patient data and the datasets received were examined for inclusion [bib_ref] Systematic reviews of healthcare: assessing the quality of controlled clinical trials, Juni [/bib_ref].
## Study endpoints
Out of the 27 studies included (conducted between 1999 and 2009), three were multi-country studies [bib_ref] Amodiaquine-artesunate versus amodiaquine for uncomplicated Plasmodium falciparum malaria in African children: a..., Adjuik [/bib_ref] [bib_ref] Randomized, multicentre assessment of the efficacy and safety of ASAQ -a fixed-dose..., Ndiaye [/bib_ref] , two studies compared ASAQ fixed and loose combination [bib_ref] The efficacy and safety of a new fixed-dose combination of amodiaquine and..., Sirima [/bib_ref] [bib_ref] Efficacy of artesunateamodiaquine and artemether-lumefantrine fixed-dose combinations for the treatment of uncomplicated..., Schramm [/bib_ref] [bib_ref] Tolerability and safety of artesunateamodiaquine and artemether-lumefantrine fixed dose combinations for the..., Schramm [/bib_ref] and three studies were non-comparative (in Sierra Leone [bib_ref] Low efficacy of the combination artesunate plus amodiaquine for uncomplicated falciparum malaria..., Grandesso [/bib_ref] and Senegal [bib_ref] Economic evaluation of a policy change from single-agent treatment for suspected malaria..., Agnamey [/bib_ref] [bib_ref] Efficacy and safety of artesunate plus amodiaquine in routine use for the..., Brasseur [/bib_ref]. Details of the studies can be found elsewhere [bib_ref] Efficacy of artesunate plus amodiaquine versus that of artemether-lumefantrine for the treatment..., Martensson [/bib_ref] [bib_ref] Efficacy, safety, and selection of molecular markers of drug resistance by two..., Djimdé [/bib_ref] [bib_ref] Combination therapy for uncomplicated falciparum malaria in Ugandan children: a randomized trial, Dorsey [/bib_ref] [bib_ref] Efficacy of anti-malarial treatment in Guinea: in vivo study of two artemisinin..., Bonnet [/bib_ref] [bib_ref] Artesunate + amodiaquine and artesunate + sulphadoxine-pyrimethamine for treatment of uncomplicated malaria..., Swarthout [/bib_ref] [bib_ref] Efficacy of three artemisinin combination therapies for the treatment of uncomplicated Plasmodium..., Van Den Broek [/bib_ref] [bib_ref] High efficacy of two artemisinin-based combinations (artesunate + amodiaquine and artemether +..., Guthmann [/bib_ref] [bib_ref] Safety and efficacy of dihydroartemisinin/ piperaquine (Artekin) for the treatment of uncomplicated..., Karema [/bib_ref] [bib_ref] Antimalarial efficacy of chloroquine, amodiaquine, sulfadoxine-pyrimethamine, and the combinations of amodiaquine +..., Guthmann [/bib_ref] [bib_ref] Malaria in the Nuba Mountains of Sudan: baseline genotypic resistance and efficacy..., Hamour [/bib_ref] [bib_ref] Efficacy of two artemisinin combination therapies for uncomplicated falciparum malaria in children..., Van Den Broek [/bib_ref] [bib_ref] Artemisinin versus nonartemisinin combination therapy for uncomplicated malaria: randomized clinical trials from..., Yeka [/bib_ref] [bib_ref] Is amodiaquine failing in Rwanda? Efficacy of amodiaquine alone and combined with..., Rwagacondo [/bib_ref] [bib_ref] Combination treatments for uncomplicated falciparum malaria in Kampala, Uganda: randomised clinical trial, Staedke [/bib_ref] [bib_ref] Artemisinin combination therapies for treatment of uncomplicated malaria in Uganda, Bukirwa [/bib_ref]. There was one unpublished study conducted in RDC (Cohuet, unpublished). The analysis of parasite clearance was by modified intent-to-treat (ITT), including all participants who were randomized and received the study medications. The primary endpoint in all the studies included in the analysis was parasitological efficacy, except one study about tolerability [bib_ref] Tolerability and safety of artesunateamodiaquine and artemether-lumefantrine fixed dose combinations for the..., Schramm [/bib_ref]. Follow-up ceased at the time of parasitological failure (either primary or recurrence), loss to follow-up, protocol violation and no data were recorded thereafter. Parasitaemia was recorded at enrolment (Day 0) and post-treatment from Day 1 through Day 28, but sampling schedules varied across studies during the first seven days, with Day 1 or Day 2 not recorded in all. Consecutive parasite slide results (Day 0, Day 1, Day 2, Day 3, Day 7) were available at 18 sites from five randomized controlled trials (RCT) [bib_ref] Amodiaquine-artesunate versus amodiaquine for uncomplicated Plasmodium falciparum malaria in African children: a..., Adjuik [/bib_ref] [bib_ref] Randomized, multicentre assessment of the efficacy and safety of ASAQ -a fixed-dose..., Ndiaye [/bib_ref] [bib_ref] Efficacy of artesunate plus amodiaquine versus that of artemether-lumefantrine for the treatment..., Martensson [/bib_ref] [bib_ref] Efficacy, safety, and selection of molecular markers of drug resistance by two..., Djimdé [/bib_ref].summarizes the measures and analyses conducted. The risks of delayed parasite clearance by Day 2 and Day 3 and parasite clearance failure were analysed as binary variables using age (continuous in years), baseline parasitaemia (log transformed) as continuous variables, anaemia (binary variable, haemoglobin < 10 g/dL), study year (continuous) and study (categorical). The parasite reduction ratio on Day (n) was defined as: parasitaemia before treatment/parasitaemia on Day (n).
## Definitions and analyses
The curve of the proportions of patients remaining parasitaemic during follow-up was fitted using logistic regression to give an estimate for clearance time. The fitted model for the logit of the proportion of parasitaemic patients over time and expressed in hours from the logistic regression:
[formula] p day ð Þ ¼ Logit constant-a à day þ 1 ð Þ ð Þ with p 0 ð Þ ¼ Logit constant ð Þ pPCT hour ð Þ¼ Logit −constant=a ð Þ -1 ð Þ Ã 24 [/formula]
In order to simplify the calculations of the predicted parasite clearance time ((pPCT), for a patient or a group of patients using the median time parasite clearance) the patient population was classed into ten pre-treatment parasite densities categories. A logistic model was fitted for each group and the corresponding predicted PCT was calculated, and once the results of predicted PCT for the ten groups were obtained, a simple logarithmic model was fitted with the aggregated results.
The risks, presented as adjusted risks ratios (AOR), were assessed by logistic multivariate analysis with random effect on the site in an attempt to account for potential statistical heterogeneity while controlling for age (continuous), parasitaemia before treatment (log-transformed), anaemia (binary), and year (continuous). Categorical data were compared using the Chi-square or the Fisher exact test as appropriate. The Spearman test was used to analyse the relationship between clearance reduction and pre-treatment parasitaemia. Confidence intervals (CI) were calculated at 95% and statistical significance was set at p-value <0.05. Modelling the time of parasite clearance Predict the time of parasite clearance using the parasitaemia before treatment and the proportions of positive patients
# Ethical issues
All studies had been approved by the relevant ethics and institution review committees as reported in the individual papers.
## Asaq treatment regimens
The majority of the patients were treated with individually formulated AS and AQ. The target dose was AS 12 mg/kg over three days and AQ 30 mg/kg over three days except in Uganda where AQ was given at 25 mg/kg (Day 0: 10 mg/kg, Day 1: 10 mg/kg, Day 2: 5 mg/kg). The loose combinations of ASAQ were dosed based on body weight, while in four studies the fixed dose combination (FDC) ASAQ was based on age and weight range [bib_ref] Randomized, multicentre assessment of the efficacy and safety of ASAQ -a fixed-dose..., Ndiaye [/bib_ref] [bib_ref] The efficacy and safety of a new fixed-dose combination of amodiaquine and..., Sirima [/bib_ref] [bib_ref] Efficacy of artesunateamodiaquine and artemether-lumefantrine fixed-dose combinations for the treatment of uncomplicated..., Schramm [/bib_ref] [bib_ref] Tolerability and safety of artesunateamodiaquine and artemether-lumefantrine fixed dose combinations for the..., Schramm [/bib_ref]. The FDC was also given either once or twice a day [bib_ref] Randomized, multicentre assessment of the efficacy and safety of ASAQ -a fixed-dose..., Ndiaye [/bib_ref]. ASAQ FDC (Coarsucam™ Winthrop® Sanofi Aventis); AS 25 mg/AQ 67.5 mg one tab/day for three days in children 5-8.9 kg; AS 50 mg/AQ 135 mg one tab/day for three days in children 9-17.9 kg; AS 100 mg/AQ 270 mg one tab/day for three days in children 18-35.9 kg.
Comparator treatment regimens i. ACT: artemether-lumefantrine (AL)(20 mg artemether/120 mg lumefantrine) given according to weight as one (5-14 kg), two (15-24 kg), three , and four (≥35 kg) tablets given twice daily co-administrated with fat for three days; Coartem™, Novartis); dihydroartemisininpiperaquine (DP) was given once daily over three days, at the standard dosage of 2.25 mg/kg and 18 mg/kg of dihydroartemisinin-piperaquine, respectively, rounded up to the nearest half tablet (two formulations were used: 20 mg dihydroartemisinin + 160 mg piperaquine and 40 mg dihydroartemisinin +20 mg piperaquine, Eurartesim™, Sigma-Tau, and in Rwanda, Artekin™, Holley); AS + sulphadoxine-pyrimethamine (SP)(AS 4 mg/kg/day; SP 25 mg/kg of sulphadoxine and 1.25 mg/kg of pyrimethamine administered in a co-formulated tablet as a single dose); ii. Non-ACT: AQ + SP (AQ 10 mg/kg/day for three days and SP 25 mg/kg of sulphadoxine and 1.25 mg/ kg of pyrimethamine administered in a coformulated tablet (SP) as a single dose); chloroquine (CQ)(25 mg/kg over three days) and SP; AQ monotherapy (10 mg/kg/day for three days); AS monotherapy (AS 12 mg/kg over five days).
# Results
Overall 15,017 patients treated for uncomplicated P. falciparum malaria at 44 sites in 20 sub-Saharan African countries were included in the analysis [fig_ref] Figure 1: Flow chart, number of patients with parasitaemia results and daily results from... [/fig_ref]. The proportion of patients treated with ASAQ was 51% (n = 7,660) vs 49% (n = 7,357) on comparator treatments; 31% (4,848/15,017) of the patients had complete parasitaemia data every 24 hours up to Day 3 and then Day 7 (2,355 in ASAQ and 3,493 in comparator groups). Overall, the proportion of children under six years of age was 77%, and the overall geometric mean parasitaemia pretreatment was 16,964/μL [fig_ref] Table 2: Patients at enrolment [/fig_ref].
## Parasitaemia on day 2 and day 3
Delayed parasite clearance, defined as the proportion of patients still parasitaemic on Day 2 or Day 3 (analysed as a binary variable) under ASAQ treatment in comparative and non-comparative trials was analysed in 44 sub-Saharan African sites over the period 1999-2009. The proportion of patients on ASAQ who were still parasitaemic on Day 2 was 8.6% (603/7,020, 95% CI 7.9-9.3%) and ranged from 1.0% in Burkina Faso-Nanoro (2008) and Congo-Kindamba (2004) to 57.1% in DRC-Boende in 2003 (Additional file 1:. Using multivariate logistic regression with random effect on site and controlling for treatment, the risk factors for a patient to remain positive on Day 2 were higher parasitaemia before treatment (AOR 2.12, 95% CI 1. The proportion of patients treated with ASAQ who were still parasitaemic on Day 3 was 1.5% (116/7,550, 95% CI 1.2-1.8%, of whom 44 in RDC-Boende), ranging from 0% in many various sites across sub-Saharan Africa to 55.9% in DRC-Boende. Using multivariate logistic analysis with random effects on sites, younger patients (AOR 0.94, 95% CI 0.90-0.98, p = 0.005) and patients with higher parasitaemia at enrolment (AOR 2.43, 95% CI 1.98-3.00, [fig_ref] Figure 2: Delayed parasitaemia [/fig_ref].
## Parasite clearance failure
Overall, patients treated with an ACT were at lower risk of parasite clearance failure by Day 7 (AOR 0.03, 95% CI 0.01-0.12, p = 0.001, stratified by site) compared to non-ACT (
## Parasite reduction ratio
Overall, in studies treating with ASAQ and recording daily parasitaemia (n = 2,355), the parasite reduction ratio (parasitaemia before treatment/parasitaemia on day (n)) was 93.9% by Day 1 and 99.9% by Day 2 [fig_ref] Table 4: Clearance reduction, artesunate/amodiaquine groups [/fig_ref]. The parasite reduction ratio ranged on Day 1 from 77.1%
in Mozambique to 99.2% in Kenya with a significant correlation between clearance reduction on Day 1 and pretreatment parasitaemia (r = 0.098, p = 0.001, Spearman test, individual data).
## Predicted time of parasite clearance
The number of patients included in RCT at 17 sites with complete parasitaemia record (every 24 hours from Day 0 to Day 3 plus Day 7) was 4,848 of whom 2,355 were treated with ASAQ and 2,493 with a comparator drug. Using logistic regression, the predicted time of parasite clearance was overall ≈ 31 hours for a median baseline parasitaemia of 27,125/μL, ranging from ≈ 19 hours for patients with parasitaemia before treatment <2,500/μL to ≈ 37 hours for patients with parasitaemia >100,000/μL.
Using the predicted PCT from the ten categories of levels of baseline parasitaemia, a logarithmic relationship was detected between the observed median baseline parasitaemia and the corresponding predicted PCT [fig_ref] Figure 3: Fitted curve of parasite clearance and intervals, artesunate/amodiaquine treatment [/fig_ref]. The fitted curve was a logarithmic relationship defined by:
[formula] pPCT ¼ 3:614 Ã ln p0 [/formula]
ð Þ -6:135; r 2 ¼ 0:94 where: pPCT ¼ predicted parasite clearance time expressed in hours; and p0 ¼ parasitaemia at enrolment pre−treatment ð Þ
The lower and upper bounds of the interval were fitted using sites where the overall proportion of parasitaemic patients were <10% on Day 2 for the lower bound and >10% for the upper bound with a coefficient of determination (r 2 ) of 0.88 for both adjustments [fig_ref] Figure 3: Fitted curve of parasite clearance and intervals, artesunate/amodiaquine treatment [/fig_ref]. The 10% threshold was in keeping with WHO recommendations.
# Discussion
This individual patient data analysis of parasite clearance was conducted on a large sample (n~15,000) of uncomplicated malaria treatments (predominantly paediatric; three-quarters less than six years old) from 44 sites and 20 sub-Saharan African countries with different levels of endemicity, covering the decade 1999-2009. This analysis shows that rapid Plasmodium falciparum clearance continues to be achieved in sub-Saharan African patients treated with ACT, and in particular with ASAQ, although direct within-site comparisons are not available; it has therefore benchmarking value as reference for future studies. This study also presents a prediction formula for PCT as a pragmatic tool adapted for studies with binary data (not quantitating parasite densities) and once-daily sampling, such as those that would be done at the typical study site in peripheral settings and in routine surveillance by control programmes. This prediction could apply to groups of patients as well as to the individual patient. These results confirm the correlation between pretreatment parasitaemia and PCT [bib_ref] In vivo parasitological measures of artemisinin susceptibility, Stepniewska [/bib_ref] , and offer an easy way to calculate the typical PCT (in hour) using parasitaemia before treatment (p0) with ASAQ or other equivalent treatments (pPCT = 3.614*ln(p0) -6.135, r 2 = 0.94). There was also a (weaker) correlation between parasitaemia before treatment and the parasite reduction ratio (PRR) at 24 h.
Monitoring treatment response and detecting artemisinin resistance as early as possible have become a major issue in malaria control. The rate at which treatment clears parasites within the first few days is at present the most useful practical test for ACT, as early response to treatment relies predominantly on the parasite response to artemisinin, independent of whether parasites are later cleared for good through the combination of the longerlived companion drug and the host's immune response.
The present analysis focused on initial response to ACT treatment, as a proxy for artemisinin resistance and failure, and aimed to identify variables that were independently associated with persistent parasitaemia on Days 2 and 3 in settings where artemisinin resistance has probably not yet emerged.
Within the period covered by these studies, there was no indication of delayed response to ACT. ACT cleared parasites rapidly leaving very few patients with a low parasitaemia on Day 3 (first day after the completion of the three-day ACT regimen). In particular, with ASAQ the parasite reduction ratio was 93.9% by Day 1 and 99.9% by Day 2; only 1.5% of the patients were still positive on Day 3 (against a proposed threshold of 3% indicating delayed response [bib_ref] In vivo parasitological measures of artemisinin susceptibility, Stepniewska [/bib_ref]. These data confirm that pre-treatment parasitaemia is a risk factor for failing to clear parasites on Day 2 and 3. They also show a clear contribution of artemisinins to the speed of action: ACT performed significantly better than non-ACT (whether combinations or single-agent treatments); no difference was detected between ACT (whether the artemisinin derivative was artesunate, artemether or dihydroartemisinin).
Anaemia was an additional risk factor for failing to clear parasites by Day 2; younger patient's age for Day 3. The only exception was the study conducted in DRC-Boende (Cohuet, unpublished) which had a high proportion of parasite clearance failures with both ASAQ and AL. These findings are difficult to explain and should be taken with caution, but one cannot exclude that there might also have been a focus of resistant Plasmodium falciparum in the DRC. No subsequent studies are available to confirm or belie this original study. Understanding the dynamics of parasite clearance could help detect early signs of artemisinin resistance and distinguish biologically meaningful changes in early parasite clearance over time from changes that may be due to the role of effect modifiers like higher pre-treatment parasitaemia, young age and anaemia by applying multivariate analysis.
Consistently with previous results [bib_ref] In vivo parasitological measures of artemisinin susceptibility, Stepniewska [/bib_ref] and the firstorder process of parasite clearance, there was a logarithmic correlation between patients' pre-treatment parasitaemia and their pPCT; while developed using ASAQ data, it was applied satisfactorily to other ACT within this database as well as other studies. For example, in a study conducted in Mali [bib_ref] No evidence of delayed parasite clearance after oral artesunate treatment of uncomplicated..., Maiga [/bib_ref] where parasitaemia was tested with frequent screenings (every eight hours) in patients treated for uncomplicated P. falciparum with artesunate for seven days and where there was no evidence of delayed parasitaemia, the median parasitaemia at enrolment (P0) was 27,070 μL and the observed median time of parasite clearance was 32 hours. Applying the suggested method using the logarithmic correlation found in the present paper (pPCT = 3.614*ln(P0) -6.135) gave very similar results (30.8 hours with intervals of 26.8 to 37.3 hours).
This simple method was tested to predict a "typical" PCT in situations when resistance has not emerged. However, this estimation of the predicted time to parasite clearance is not without its limitations. The calculations may lack precision with studies sampling once a day and at variable intervals from treatment intake (as it might be the case for most clinical trials, especially on the first day). It has been suggested that parasite clearance rate can only be accurately estimated if sampling is at least every six hours [bib_ref] The parasite clearance curve, White [/bib_ref] ; it is thus possible that data could have been "over-analysed" in generating predicted time of parasite clearance in this case. Although parasite clearance half-life is the research reference to define slow parasite clearance, sampling patients four times or even twice a day is not feasible in routine monitoring, and might be cumbersome even in research settings. Detecting the first signs of artemisinin resistance in uncomplicated P. falciparum will likely require a combination of more sophisticated studies and routine monitoring. Between parasite clearance estimators [bib_ref] Standardizing the measurement of parasite clearance in falciparum malaria: the parasite clearance..., Flegg [/bib_ref] and Day 3 persistency [bib_ref] In vivo parasitological measures of artemisinin susceptibility, Stepniewska [/bib_ref] , simple estimations of PCT based on daily sampling may add elements to facilitate comparisons and detection of trends. It would be useful to compare methods for validation, including databases with delayed responses.
[fig] Figure 1: Flow chart, number of patients with parasitaemia results and daily results from Day 0 to Day 3 and Day 7. [/fig]
[fig] Figure 2: Delayed parasitaemia (Day 3) forest plot, artesunate/ amodiaquine vs comparator treatments. AL: artemetherlumefantrine; AOR: adjusted odds ratio using multivariate logistic regression with random effects; AQ: amodiaquine; AS: artesunate; CQ: chloroquine; DP: dihydroartemisinin-piperaquine; fdc: fixed dose combination; na: not applicable; SP: sulphadoxine-pyrimethamine. [/fig]
[fig] Figure 3: Fitted curve of parasite clearance and intervals, artesunate/amodiaquine treatment. A: pPCT, predicted parasite clearance time. B: pPCT, predicted parasite clearance time. [/fig]
[table] Table 2: Patients at enrolment [/table]
[table] Table 3: Proportion of patients with a parasite clearance failure by Day 7 and by treatment groups [/table]
[table] Table 4: Clearance reduction, artesunate/amodiaquine groups [/table]
|
Personalized Reminders for Immunization Using Short Messaging Systems to Improve Human Papillomavirus Vaccination Series Completion: Parallel-Group Randomized Trial
Background: Completion rates among adolescents who initiate the human papillomavirus (HPV) vaccine 3-dose series are low. SMS text message vaccine reminders are effective, but less is known about the best types for HPV series completion or the ability to assess and target vaccine decision-making stage.Objective: The aim of this study is to compare the effectiveness of HPV vaccine series completion in minority adolescents who received precision and educational versus conventional SMS text message reminders.Methods: Enrolled parents of adolescents aged 9-17 years who received the first HPV vaccine dose at 1 of the 4 academic-affiliated community health clinics in New York City were randomized 1:1 to 1 of the 2 parallel, unblinded arms: precision SMS text messages (which included stage-targeted educational information, next dose due date, and site-specific walk-in hours) or conventional SMS text messages without educational information. Randomization was stratified according to gender, age, and language. The primary outcome was series completion within 12 months. In post hoc analysis, enrollees were compared with concurrent nonenrollees and historical controls.Results:Overall, 956 parents were enrolled in the study. The precision (475 families) and conventional (481 families) SMS text message arms had similarly high series completion rates (344/475, 72.4% vs 364/481, 75.7%). A total of 42 days after the first dose, two-thirds of families, not initially in the preparation stage, moved to preparation or vaccinated stage. Those in either SMS text message arm had significantly higher completion rates than nonenrollees (708/1503, 47.1% vs 679/1503, 45.17%; P<.001). Even after removing those needing only 2 HPV doses, adolescents receiving any SMS text messages had higher completion rates than historical controls (337/2823, 11.93% vs 981/2823, 34.75%; P<.001). A population-wide effect was seen from 2014 to 2016, above historical trends.Conclusions: SMS text message reminders led to timely HPV vaccine series completion in a low-income, urban, minority study population and also led to population-wide effects. Educational information did not provide an added benefit to this population.
# Introduction
Background Adolescents, particularly minority adolescents, are not adequately protected against human papillomavirus (HPV) and its potential sequelae, including cancer and genital warts. Despite the highly efficacious vaccine being recommended for all adolescents, completion rates among those who initiate the series are low [bib_ref] National, regional, state, and selected local area vaccination coverage among adolescents aged..., Elam-Evans [/bib_ref]. Nationwide, only 71.5% of adolescents aged 13-17 years initiate the HPV vaccine series, and half (54.2%) have received all needed doses [bib_ref] National, regional, state, and selected local area vaccination coverage among adolescents aged..., Elam-Evans [/bib_ref]. Adherence to the recommended HPV vaccine dosing schedule is also exceedingly low; one study found that of 9-to-16-year-olds who had initiated HPV vaccination, only 28% completed the then-recommended 3-dose series within 1 year [bib_ref] Adherence to the HPV vaccine dosing intervals and factors associated with completion..., Widdice [/bib_ref]. Caregiver-decided vaccination delays can significantly contribute to the spread of infectious diseases in adolescents [bib_ref] Parental decision-making on childhood vaccination, Damnjanović [/bib_ref]. This is a particularly salient factor to counter for HPV vaccination, as HPV infection not only carries short-term infection risk but also long-term chronic disease risk.
Health information technology interventions that link communication technologies, such as SMS text messaging, with electronic health record data offer low-cost, scalable opportunities to foster vaccination and other preventive care behaviors [bib_ref] Scoping review and evaluation of SMS/text messaging platforms for mHealth projects or..., Iribarren [/bib_ref]. Currently, there is a growing body of work supporting the feasibility, acceptability, and efficacy of eHealth and mobile health (mHealth) technologies [bib_ref] Behavioral functionality of mobile apps in health interventions: a systematic review of..., Payne [/bib_ref] [bib_ref] Can mobile phone apps influence people's health behavior change? An evidence review, Zhao [/bib_ref] , particularly within the realm of adolescent and child health promotion [bib_ref] Patient-centered ehealth interventions for children, adolescents, and adults with sickle cell disease:..., Badawy [/bib_ref] [bib_ref] eHealth and mHealth interventions in pediatric cancer: a systematic review of interventions..., Ramsey [/bib_ref].
Studies have demonstrated the effectiveness of mHealth SMS text message interventions on vaccination coverage and timeliness at levels in line with other forms of reminder or recall, particularly in low-income populations for whom other forms of reminder recall have been less successful [bib_ref] Text messaging immunization reminders: feasibility of implementation with low-income parents, Ahlers-Schmidt [/bib_ref] [bib_ref] Effect of a text messaging intervention on influenza vaccination in an urban,..., Stockwell [/bib_ref] [bib_ref] Effectiveness and cost of bidirectional text messaging for adolescent vaccines and well..., O'leary [/bib_ref] [bib_ref] Patient reminder and recall interventions to improve immunization rates, Jacobson [/bib_ref] [bib_ref] Text message reminders to promote human papillomavirus vaccination, Kharbanda [/bib_ref]. SMS text message reminders have been found to increase HPV vaccine uptake in various populations, particularly in the adolescent cohort [bib_ref] Effectiveness and cost of bidirectional text messaging for adolescent vaccines and well..., O'leary [/bib_ref] [bib_ref] Understanding the use of digital technology to promote human papillomavirus vaccination -A..., Stephens [/bib_ref] [bib_ref] Effectiveness of centralized text message reminders on human papillomavirus immunization coverage for..., Rand [/bib_ref] [bib_ref] Parental choice of recall method for HPV vaccination: a pragmatic trial, Kempe [/bib_ref] [bib_ref] Parental education and text messaging reminders as effective community based tools to..., Aragones [/bib_ref] [bib_ref] Comparison of reminder methods in selected adolescents with records in an immunization..., Morris [/bib_ref] [bib_ref] Effects of phone and text message reminders on completion of the human..., Rand [/bib_ref] [bib_ref] Direct messaging to parents/guardians to improve adolescent immunizations, Bar-Shain [/bib_ref] [bib_ref] Short message service reminders to parents for increasing adolescent human papillomavirus vaccination..., Tull [/bib_ref]. However, their effect has not been as robust as needed. One potential advantage of SMS text message interventions, which has not been well investigated, is the ability to provide precision messages. Such tailored messages may promote increased engagement with SMS text message reminders and, in turn, positively impact HPV vaccination completion rates.
## Objectives
In this study, we compare the impact of precision SMS text message on HPV vaccine series completion (tailored vaccine health literacy-promoting information targeted to the family's stage of vaccine decision-making) with conventional SMS text message reminders in a pragmatic randomized trial. The transtheoretical model of behavior change guided the tailoring of SMS text messages.
# Methods
## Overview
This trial was conducted in 4 community health clinics affiliated with the New York-Presbyterian Hospital Ambulatory Care Network and Columbia University between December 2014 and December 2017. These practices provide care for primarily publicly insured Latino patients. The Vaccines for Children program provides free vaccines for nearly all the patients at the study sites, and all the study sites allow walk-ins for second and third HPV vaccine doses without an appointment. All vaccinations given at the study sites are documented in the New York-Presbyterian Hospital immunization registry, which extracts information about vaccinations directly from the provider order entry module of the electronic health record, making data accurate for HPV vaccines administered at clinical sites. The registry also synchronizes data with the New York Citywide Immunization Registry (CIR), which is a population-based registry. The New York City Public Health Law requires documentation for all vaccinations administered to those aged ≤18 years old be submitted to CIR, which captures >85% of vaccines administered in New York City and 93% of vaccines from the Vaccines for Children program.
Recruitment followed a 2-pronged enrollment process. First, nurses at the study sites provided families with a recruitment card and information sheet. Families interested in being contacted could put a cell phone number on the card, which was then left for the research assistants. Those who did not want to be contacted could also indicate this as such. Of the 547 families from whom a recruitment card was collected, only 9 (1.6%) indicated that they wished not to be contacted. Permission was received to contact the families of all adolescents who received their first HPV vaccine dose at one of the study sites during the enrollment period for whom a card was not left or for whom the nurse did not have time to give a card to assess eligibility and interest.
Parents or legal guardians involved in the study had to meet the following eligibility criteria: (1) have a child aged between 9 and 17 years who received their first HPV vaccine at study clinics, (2) own a cell phone with SMS text message capabilities, (3) have English or Spanish literacy, (4) plan to remain in New York City for the next 12 months, and (5) have not been previously contacted to enroll with a different child. Children down to 9 years were included as the vaccine was licensed down to those aged 9 years. There was no compensation for enrollment in the study.
After each enrollment, the project coordinator used an adaptive electronic randomization algorithm to parallel randomize all participants with a 1:1 allocation ratio stratified by (1) clinic site, (2) adolescent's gender, (3) adolescent age group (9-14 years vs 15-17 years), and (4) parental language. The scheme was adaptive in that the assignment of each family was to the arm that would keep the allocation ratios in the 4 strata the family belonged to (site, sex, age, and language) closer to 1:1. The statistician and analyst were blinded to the arm assignments.
We designed the messages (precision and conventional) using information gathered during our previous studies, the relevant literature, and expertise in SMS text messaging, HPV vaccination, adolescent medicine, health literacy, and the community. Messages were then pretested with 20 parents iteratively until no new message changes were made. Parents participating in this pretest received a round trip New York City Metrocard.
For families receiving precision SMS text message reminders, an automated, in-house SMS text messaging software program first used a short cascade of questions based on the transtheoretical model to assess the family's stage of decision-making regarding vaccination [fig_ref] Figure 1: Transtheoretical model stage allocation in the day 21 message [/fig_ref]. This was sent on day 21 after administration of the first HPV vaccine dose. On the basis of the person's response, the platform automatically placed them into the correct stage and proceeded to send educational information targeted at that stage of decision-making. Parents who were in the precontemplation stage were unaware that their adolescents needed a second vaccination or when it was due. Those in contemplation knew their child or adolescent needed a dose but might still have had questions such as vaccine efficacy, side effects, and safety. Finally, those in preparation were planning to have their adolescents continue the vaccination series but might not have known where or when to access care. The program was monitored by the project coordinator. Parents in each stage received different messages [fig_ref] Figure 2: Precision and conventional arm SMS text message examples [/fig_ref]. For example, messages for parents in precontemplation first notified them that their adolescents were in need of subsequent doses and why those doses were needed, whereas messages for parents in the contemplation stage provided information needed to answer any remaining questions they might have had regarding vaccination. For some messages were responsive to user input, such that parents were able to self-tailor the content by texting back indicators for which items they wanted more information about. For those in the preparation stage, the messages provided information regarding where and when to walk-in for vaccination. Parents in the other 2 stages also received information regarding where and when to walk-in for vaccination. Families also had 2 additional instances, on days 33 and 40, where their current stage was assessed, if they were not already in the preparation stage. On the basis of responses to these messages, families could switch into a different stage track of messages.
## Conventional sms text message reminders
Parents not randomized to the precision text message arm received conventional text reminders notifying them when the next dose was due. These messages did not include vaccine health literacy-promoting information and were similar to those used in our previous adolescent studies [fig_ref] Figure 2: Precision and conventional arm SMS text message examples [/fig_ref].
## Message frequency
Informational messages for both arms began on day 28. This was designed so that, if a family reacted to the message and came in to be vaccinated, it would not be before the 28-day minimal required interval between the first and second dose; the intervention began before the HPV recommendation changed from 3 to 2 doses for younger populations. Subsequent messages were sent on days 35, 42, 49, and 56 post vaccination for both study arms. These dates were chosen because of their proximity to the initial vaccination date; the second dose was due on day 56 (counting from the day the first dose was administered). We selected reminder message send dates at the time they were due and included an additional 1 week (63 days), 2 weeks (70 days), 4 weeks (84 days), and 6 weeks (98 days) overdue reminders. This series of 5 messages (days 28, 35, 42, 49, and 56) was selected based on the protocol from our previous trial in which a median of 5 messages was needed for a family to bring a child in for vaccination and was well tolerated by parents with very few stop requests [bib_ref] Effect of a text messaging intervention on influenza vaccination in an urban,..., Stockwell [/bib_ref]. Booster messages were also sent on days 63, 70, 84, and 98 post vaccination. Once or if an adolescent received the second dose, parents began receiving their next set of messages 28 days before the due date of the next dose, and then followed the same timing as with the second dose messages (days 28, 21, 14, 7, and 0 before the third dose).
Messages were sent in English or Spanish, based on parent preference. On the basis of our previous study [bib_ref] Parental and provider preferences and concerns regarding text message reminder/recall for early..., Hofstetter [/bib_ref] , which identified parental preferences, messages were designed to include the child's name and stating that it was being sent on behalf of the clinic. The recipient of the messages was the parent, rather than the child, based on previous preferences elicited in this population. Families stopped receiving messages on their original schedule when the required dose was abstracted from the hospital's immunization registry, which included synchronized data from the CIR, as described earlier.
In October 2016, the Centers for Disease Control and Prevention (CDC) Advisory Committee on Immunization Practices recommended a change to 2 doses of the HPV vaccine series (0 and 6 months) for those children or adolescents who start the series under the age of 15 years. There was no change in those who started the series when they were aged ≥15 years. In response, the following alterations were made: parents of adolescents <15 years when first vaccinated, who were already enrolled, and had not yet received their second dose, received an SMS text message update informing them that their child now only needed 2 doses 6 months apart and that we would text them when the second dose was due. These parents then received an updated series of messages consistent with the new 2-dose recommendation in terms of both phrasing and timing. Parents of adolescents who either initiated the series (1) at the age of ≥15 years or (2) at <than 15 years but had already received their second dose less than 6 months after the first, stayed on the original protocol because the adolescent was still in need of 3 doses. Any parents of already-enrolled adolescents who were aged <15 years when first vaccinated and received their second dose at least 5 months after their first were considered to have completed the series under the new recommendations. These parents received an SMS text message letting them know that, because their child's second shot was at least 5 months after the first, they had now completed the series.
# Measures and analysis
On the basis of previous baseline data, 50.9% (245/481) of patients in the conventional group were expected to receive 3 doses in 12 months. With a sample size of 956, we were powered at 80% to detect a 9% difference in completion rate (the primary outcome) between arms to be statistically significant at P=.05. The primary outcome measure was timely HPV vaccine series completion within 12 months (operationalized as the receipt of 2 or 3 doses, based on age and enrollment date, and accounting for the 2016 CDC guideline change). Vaccine completion was extracted from local vaccine sources.
HPV vaccine completion rates were compared for all adolescents of participant parents at the end of a 12-month observation period starting at the receipt of their first HPV vaccine dose. All primary analyses were conducted on an intention-to-treat basis. Completion rates in the 2 randomized groups were compared using 2-tailed chi-squared tests at a significance level of P<.05. Multiple logistic regressions were performed to assess any potential differences in receipt among demographic groups.
Chi-square tests were conducted as a post hoc analysis to compare both intervention arms with concurrent nonenrollees (n=1503) who received their first vaccine dose during the study period, as well as with historical controls (n=2823; first dose administered 2011-2013). Intervention-arm adolescents who received 2 doses per the new guideline were removed to achieve comparability for the historical analysis. This study was approved by the Columbia University Irving Medical Center Institutional Review Board. Personalized Reminders for Immunization Using Short Messaging is registered with ClinicalTrials.gov with identifier NCT02236273.
## Role of the funding source
The Agency for Healthcare Research and Quality funder of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.
# Results
## Overview
In total, we screened 1593 adolescents who received their first HPV vaccine dose at the study sites. The majority (1454/1593, 91.3%) were able to be contacted. Of the contacted patients, only 2% (29/1454) of families did not have a cell phone with SMS text messaging, and 11.1% (161/1454) of were excluded based on other exclusion criteria. Of the eligible families, most (956/1264, 75.63%) were enrolled .
## Movement through stages
Overall, 12,000 messages were sent. Of those randomized to the intervention arm, 1 family received their second dose early, and therefore, did not receive second dose messages; a second family requested to stop the program before the messages started. This left 473 of the 475 families randomized to the intervention arm eligible to receive messages. Most families (428/473, 90.5%) in the precision reminder arm received the day 21 message. There were technical issues for 45 families (45/473, 9.5%), as some messages were undelivered because of user service disruption. Of those who received the messages, two-thirds (288/428, 67.3%) of families responded to stage-assessment messages: 52.6% (225/428) were in preparation, [bib_ref] Effect of a text messaging intervention on influenza vaccination in an urban,..., Stockwell [/bib_ref] The movement was similar for the third dose. Overall, 336 families randomized to the precision message arm needed a third dose of the vaccine, including those for whom a third dose was needed based on the year and age at first dose. We received replies from half (181/336, 53.9%) of the participants. Of those who were not in the preparation stage at the beginning of the third dose cycle, half were able to be moved into preparation within 2 weeks before the next dose was due.
## Receipt of hpv vaccination
Of those receiving SMS text messages, Spanish-speaking parents (adjusted rate ratio 1.17, 95% CI 1.08-1.27) had an increased rate of timely completion; ≥15 years old had a decreased completion rate (adjusted rate ratio 0.95, 95% CI 0.93-0.98). No differences existed in terms of sex, education, or insurance.
Overall, both the precision SMS text message (344/473, 72.7%) and conventional (364/481, 75.7%) arms had very high timely series completion rates within 12 months, which did not significantly differ (P=.25). We found a significant difference in completion rate for those who responded to the first day 21 intervention messages (n=291) versus those who did not respond (n=153; 219/291, 75.3% vs 100/153, 65.4%; P=.04).
In a post hoc analysis, those in either SMS text message arm had a significantly higher completion rate than the nonenrollees. In addition, even after removing those who only needed 2 doses to complete the series, those in the SMS text messaging arms had higher rates than the historical controls (n=2823 [bib_ref] Designing patient-centered text messaging interventions for increasing physical activity among participants with..., Horner [/bib_ref]. Ultimately, a population-wide effect on HPV vaccination series completion was seen during the years of the study 2014-2016, above historical trends [fig_ref] Figure 4: Human papillomavirus vaccine series completion within 12 months of initiation by year... [/fig_ref].
# Discussion
## Principal findings
In this study, the addition of educational information in SMS text messages targeted to the stage of caregiver vaccine decision-making did not provide additional benefits in this low-income, urban, minority population. However, those receiving SMS text message reminders had more timely series completion than historical controls and nonenrollees. SMS text messaging also led to population-level effects that far exceeded historical trends, illustrating the potential impact of such reminders if implemented. Although ad hoc analyses, the very large differences in completion between enrollees and nonenrollees as well as historical controls adds further support. Together, our findings suggest that SMS text message reminders could be used to increase the likelihood that caregivers will follow through with HPV vaccination series completion and could also be used to combat vaccination delay.
Of those in the intervention group, more individuals who responded to any message on day 21 received the required doses than those who did not. However, in this study population, education information in the SMS text messages did not provide additional benefits over conventional SMS text messages without educational information. Studies have shown that prolonged exposure to similarly themed messages may lead to reactance or active resistance against the health behavior the health message advocates [bib_ref] How message fatigue toward health messages leads to ineffective persuasive outcomes: examining..., Kim [/bib_ref] [bib_ref] Message fatigue: conceptual definition, operationalization, and correlates, So [/bib_ref]. In addition, for parents with an unfavorable attitude toward vaccines, educational interventions have been found to reduce the intention to vaccinate [bib_ref] Effective messages in vaccine promotion: a randomized trial, Nyhan [/bib_ref]. For the subset of our patient population who failed to interact with the SMS text messages, additional educational messages may have acted as a mental deterrent to bringing in their children for subsequent doses. It is possible that for certain populations, changing to a different modality is needed. These messages were based on formative work and pretesting with patients and their families, which is critical in the development of mHealth interventions [bib_ref] Medication adherence and technology-based interventions for adolescents with chronic health conditions: a..., Badawy [/bib_ref] [bib_ref] Technology access and smartphone app preferences for medication adherence in adolescents and..., Badawy [/bib_ref] [bib_ref] Conceptualising engagement with digital behaviour change interventions: a systematic review using principles..., Perski [/bib_ref] [bib_ref] Smokers' and drinkers' choice of smartphone applications and expectations of engagement: a..., Perski [/bib_ref] ; however, future work should also potentially include an intentional exploration with the target populations of unintended impacts of messages.
Despite this finding, our study demonstrates how SMS text message reminders could interrupt a common pathway to vaccination delay and vaccine series incompletion. Vaccination delay, particularly for HPV vaccination, is often studied as an active decision by caregivers; however, given the increase in timeliness and the lack of impact of targeted vaccine-readiness information shows that delay is often not an active decision but rather a circumstantial effect based on other factors (eg, forgetfulness) that can be mitigated with timely reminders. SMS text message reminders work as a call to action and help prompt caregivers that would otherwise vaccinate their children but may fail to bring their child back for vaccination because of other factors. Further research could be conducted to explore the benefit of this technology for direct youth use, as an increasing number of health interventions are targeted at adolescent self-use to encourage increases in their health autonomy [bib_ref] Technology use for adolescent health and wellness, Radovic [/bib_ref].
In this study, we also demonstrated the ability to use SMS text messages to assess a family's stage of vaccine decision-making and move them along the stages of the transtheoretical model. Most families (409/473, 86.5%) responded to at least 1 message prompt, and two-thirds of families who were not in preparation at the first assessment were either in preparation or vaccinated by 42 days after the first dose. Similarly, half of the families of adolescents who needed 3 doses who were not in preparation at the first assessment for that dose were either in preparation or vaccinated by the time the third dose was due. Although ultimately such an in-depth, precision intervention may not have been needed for this population, it does lay the foundation for using SMS text messaging both to assess a person's stage of decision-making and to intervene to move them through to a possible behavior change. Several SMS text messaging interventions that have targeted stages of change have been tested and found to be effective in encouraging health behavior change, namely with physical activity [bib_ref] The intervention effect of SMS delivery on Chinese adolescent's physical activity, Lau [/bib_ref] [bib_ref] Designing patient-centered text messaging interventions for increasing physical activity among participants with..., Horner [/bib_ref] [bib_ref] Text to move: a randomized controlled trial of a text-messaging program to..., Agboola [/bib_ref] , smoking cessation [bib_ref] Designing and validation of text messages for m-Health intervention for tobacco cessation..., Chahar [/bib_ref] [bib_ref] Effects of a smoking cessation program including telephone counseling and text messaging..., Kong [/bib_ref] , and diabetes care management. However, many of these studies were conducted internationally, and none have addressed HPV vaccine uptake. Our study contributes to the growing body of knowledge on SMS text messages targeted to the stage of change and presents a novel understanding of SMS text message efficacy in increasing HPV vaccination completion in adolescents. These findings may be particularly applicable given the increased levels of vaccine hesitancy in caregivers in the wake of acute COVID-19 activity. Although app-based interventions offer a number of benefits, they require users to have higher levels of technological literacy than SMS text message-based interventions. SMS text messages are preset on mobile phones and require virtually no instruction for use when receiving messages outside general literacy. The results of this study, along with prior vaccine SMS text message reminder research, underscore the sustained role SMS text messaging can still play in providing a digital precision SMS text message health approach to behavioral interventions, even in the modern mobile use landscape.
During this study, the CDC changed their recommendations for the number of doses of HPV vaccine needed by adolescents who initiated the series before the age of 15 years. An unintended benefit of this study was the demonstration of the ability of SMS text messages to facilitate rapid communication with families to inform them of the CDC schedule changes. Such ability extends the possibilities for health care providers to notify families when they need them to either take or not take a certain action. If sites had to call families to tell them not to come in, it would have required extensive personnel time. Conversely, it would have been frustrating to families if they had not been notified and had showed up too early. This real-time notification can be beneficial for both health care providers and public health practitioners. Investigating modalities and best practices of remote pediatric clinic communication with caregivers is particularly needed as we rethink health care communication and adolescent care in the wake of the height of the COVID-19 pandemic.
# Limitations
Our study has several limitations. This study took place in a single medical system that serves a primarily low-income, Latino, urban community, who may be particularly sensitive to SMS text message interventions. These findings may not be generalizable to other settings. Recommendations also changed during the intervention period. Most of the study population received the first dose before the new CDC guidelines were implemented, when 3 doses were needed. However, these were accounted for in the analyses. In addition, vaccine administration has been underreported. However, all administered vaccines are documented in the electronic health record, including synchronization with the New York CIR, which has an excellent capture rate. Therefore, underreporting of vaccinations is likely low; underreporting would also have affected the intervention and usual care groups similarly.
# Conclusions
Despite these aspects of the study, our findings lend strength to the growing body of evidence showing that mHealth or eHealth interventions such as SMS text message reminders can be used to tangibly promote child and adolescent health, particularly in the realm of HPV vaccination, in which outcomes are consistently substandard to national goals. We also demonstrate the efficacy of SMS text message reminders in a low-income, tight knit, and connected minority community, which helps answer the call to improve upon digital approaches that "address disparities in access to care related to race and ethnicity, socioeconomic status". SMS text message reminders function as an accessible, easy-to-use, low-cost remote intervention that can be rapidly deployed, although information detailing the economic impact and cost-effectiveness of these interventions should be evaluated in future studies. SMS text message reminders led to timely HPV vaccine series completion in our study population, which led to population-wide effects. Although education information did not provide added benefit in this very responsive population, we also demonstrated the feasibility of using SMS text messages to both identify a family's stage of vaccine decision-making, move them further down the pathway to behavior change, and possibly decrease HPV vaccine delay. In the face of health information changes, SMS text messages also helped facilitate real-time and remote communication of these changes to caregivers, which is needed in our post pandemic clinical pediatric landscape.
[fig] Figure 1: Transtheoretical model stage allocation in the day 21 message. HPV: human papillomavirus. [/fig]
[fig] Figure 2: Precision and conventional arm SMS text message examples. HPV: human papillomavirus. [/fig]
[fig] Figure 4: Human papillomavirus vaccine series completion within 12 months of initiation by year of initiation. [/fig]
[table] Table 1: The arms were well-balanced; 481 families were randomized to the usual care arm and 475 to the intervention arm. Most of the adolescents in the enrolled families were aged ≤14 years (880/956, 92.1%). Half of them (478/956, 50%) were female, and most of them (903/956, 94.5%) were publicly insured. Two-thirds (614/956, 64.2%) of parents/caregivers were primarily Spanish speaking, and 59.9% (573/956) had a high school education or less (Table 1). JMIR Mhealth Uhealth 2021 | vol. 9 | iss. 12 | e26356 | p. 5 https://mhealth.jmir.org/2021/12/e26356 (page number not for citation purposes) Characteristics of the study participants. [/table]
|
Realistic fear of cervical cancer risk in Japan depending on birth year
Objective: In Japan, the possible adverse events upon HPV vaccination was widely reported in the media. MHLW announced the suspension of aggressively encouraging HPV vaccination in 2013, and inoculation rate has sharply declined. The aim of the present study was estimation of future cervical cancer risk.Methods: The latest data on vaccination rate at each age in Sakai City were first investigated. The rate of experiencing sexual intercourse at the age of 12, 13, 14, 15, 16, 17 and throughout lifetime is assumed to be 0%, 1%, 2%, 5%, 15%, 25%, and 85% respectively. The cervical cancer risk was regarded to be proportional to the relative risk of HPV infection over the lifetime. The risk in those born in 1993 whom HPV vaccination was not available yet for was defined to be 1.0000.Results: The cumulative vaccination rates were 65
# Introduction
Worldwide, the frequency and the death rate of cervical cancer are the fourth highest among female cancers, 1 with estimated 530,000 people newly contracting this disease in 2012, making up 7.5% of cancer deaths among women.In the United States, the estimated number of new cases of cervical cancer in 2016 is 12,900, with 4,120 of these cases leading to death.Most cervical cancers are known to be caused by persistent infection with human papillomavirus (HPV) including HPV types 16 and 18. While cervical cancers caused by HPV-16 and 18 infection are preventable with vaccination, the inoculation rate varies depending on the country [bib_ref] Global estimates of human papillomavirus vaccination coverage by region and income level:..., Bruni [/bib_ref] and remains low despite recommendation by the Centers for Disease Control and Prevention (CDC) in the United States. With regard to this, the American Society of Clinical Oncology (ASCO) has announced that the ASCO strongly supports efforts to markedly increase the proportion of boys and girls receiving the HPV vaccine in the United States and worldwide. [bib_ref] American society of clinical oncology statement: human papillomavirus vaccination for cancer prevention, Bailey [/bib_ref] According to several studies, the obstacles to vaccination involve multiple possible factors, such as safety concerns about the vaccine, a lack of clear recommendations from health care providers, etc. [bib_ref] American society of clinical oncology statement: human papillomavirus vaccination for cancer prevention, Bailey [/bib_ref] The issue of adverse events, although not markedly different from those of other vaccines, has made this problem worse worldwide. [bib_ref] Tracking the global spread of vaccine sentiments: the global response to Japan's..., Larson [/bib_ref] In Japan, the HPV vaccine was approved in 2009, after which an urgent promotion project for vaccination was initiated by the Ministry of Health, Labor and Welfare (MHLW) in 2010. HPV vaccinations were provided free of charge or at low cost for 13 to 16-year-old girls. From April 2013, periodical vaccination of 12 to 16-year-old girls was initiated, typically for 12-year-old girls. However, the occurrence of so-called serious adverse events upon HPV vaccination after initiating periodical vaccinations was widely reported in the media and a picture of a patient who exhibited symptoms was repeatedly broadcast. As a consequence, MHLW decided for the "suspension of aggressively encouraging vaccination" in June 2013.As a result of this determination, the inoculation rate has sharply declined. [bib_ref] Japan's failure to vaccinate girls against human papillomavirus, Ueda [/bib_ref] Hanley et al. reported that, after the suspension of the governmental recommendation, the completion rate of HPV vaccination plummeted to just 0.6% in Sapporo, Japan in 2014 9 . Sekine et al. reported that the vaccination rates of the 12 to 16-year-olds in Niigata City, Japan, in 2014 and 2015 were 3.5% and 0.19% respectively. [bib_ref] Japanese Crisis of HPV Vaccination, Sekine [/bib_ref] The news of Japan's suspension of the HPV vaccine recommendation has traveled globally through online media and social media networks, being applauded by anti-vaccination groups but not by the global scientific community. Larson HJ described that the longer the uncertainty around the Japanese HPV vaccine recommendation persists, the further the public concerns are likely to travel. [bib_ref] Tracking the global spread of vaccine sentiments: the global response to Japan's..., Larson [/bib_ref] The World Health Organization (WHO) stated that policy decisions based on weak evidence, leading to lack of use of safe and effective vaccines, can result in real harm.Cervical cancer patients number 8,000 people a year in Japan, and 3,500 die per year. [bib_ref] Japan Cancer Surveillance Research Group : Cancer incidence and incidence rates in..., Matsuda [/bib_ref] Furthermore, a marked increase in the incidence rate of cervical cancer in 20 to 49-year-olds has been seen. [bib_ref] Japan Cancer Surveillance Research Group : Cancer incidence and incidence rates in..., Matsuda [/bib_ref] [bib_ref] Worldwide trends in cervical cancer incidence: impact of screening against changes in..., Vaccarella [/bib_ref] This is assumed to be a result of a decline in the age of initial incidence of HPV infection due to a decline in the age of first sexual intercourse, in addition to low screening rate. Among 22 developed countries, the cervical cancer screening rate in Japan is the lowest at 37.7%, compared with that of 85.0% in the United States and more than 65.0% in all European countries.A greater decline in cervical cancer screening rate is observed in young women and is 10.2% in 20 to 25-year-olds in Japan.Although the WHO points out that the benefit of HPV vaccine is particularly great in countries having a low cervical cancer screening rate.We previously predicted that if aggressive encouragement of vaccination were not resumed going forward, the significant difference in the HPV infection rate would be observed according to the year of birth, and it would vary significantly in accordance with the year of restart of the governmental recommendation, using an assumed inoculation rate. [bib_ref] Outcomes for girls without HPV vaccination in Japan, Tanaka [/bib_ref] In the previous study, all female children were hypothesized to be vaccinated at the youngest targeted age. However, all the girls did not undergo vaccination at the youngest targeted age in practice. Moreover, the most serious public and medical concern should be future cervical cancer risk caused by infection of HPV without vaccination. The lifetime risk of cervical cancer does not correlate with the risk of HPV-16 and 18 infection at the age of 20. The aim of the present study was estimation of the lifetime cervical cancer risk by the year of birth, based on the latest real data on vaccination rate. shows the vaccination rate and the resulting change in cervical cancer relative risk that could actually occur in the future and was already being determined, confirming that our previous prediction is now becoming a reality and severity. The future cervical cancer risk in those born in 1993 was defined to be 1.0000. As a result, it was found that the relative risk in those born in 1994 was 0.70, that in those born in 1995 was 0.63, that in those born in 1996 was 0.59, that in those born in 1997 was 0.56, that in those born in 1998 was 0.56, that in those born in 1999 was 0.60, that in those born in 2000 was 0.98, that in those born in 2001 was 0.99, that in those born in 2002 was 1.0, and that in those born in 2003 was 1.0. It was clear that those born between 1994 and 1999 were part of the generation in which the prevention of future cervical cancer by HPV vaccine would be achievable to some extent, while those born after 2000 would be exposed to the same risk as before introduction of the vaccine. For these females, cervical cancer screening should be recommended more strongly.
# Results and discussion
The real inoculation rates showed how the reporting of possible adverse events in the media and the suspension of encouragement by MHLW influenced girls of the targeted age. From the above, at present, it is confirmed that the cumulative initial vaccination rates were different by the year of birth, leading to the realistic fear of differences in the risk of developing cervical cancer depending on the year of birth. It is a very unfortunate prediction that the different risks of cervical cancer depending on the year of birth will be caused by the governmental decision. The change in cervical cancer relative risk could actually occur in the future and was already being determined. In fact, Ozawa et al. demonstrated the significant reduction of the rate of abnormal cervical cytology among the vaccinated women aged 20-24 compared with those who had not received HPV vaccine in Japan. [bib_ref] Beneficial Effects of human papillomavirus vaccine for prevention of cervical abnormalities in..., Ozawa [/bib_ref] It is necessary for us to face the reality that the risk of the HPV infection and cervical cancer increases every day for girls who refrain from vaccination and begin to have sexual intercourse without it. For these females, cervical cancer screening should be recommended more strongly. In regard to HPV vaccine, catch-up vaccinations for females who missed the opportunity of HPV vaccination due to the suspension of the governmental recommendation, and use of 9-valent HPV vaccines (currently, only bi-and quadrivalent HPV vaccines are licensed in Japan) should be considered. [bib_ref] Struggles within Japan's national HPV vaccination: a proposal for future strategy, Tanaka [/bib_ref] There are some limitations in the present study. We used the data from Sakai City in this study; however, the vaccination situation will vary depending on the local government, and it appears that there are local governments which face a much worse reality than Sakai City. Moreover, we hypothesized that HPV infection risk is proportional to the rate of having sexual intercourse without HPV vaccination at every age and over the lifetime, and that the relative risk of cervical cancer could be determined by the relative risk of HPV infection over the lifetime. HPV infection risk is not regulated only by the rate of experiencing of sexual intercourse without vaccination, and cervical cancer risk is different by the type of HPV. Uncertainty of these hypotheses was minimized by demonstrating the relative risk compared with the risk of those born in 1993 whom HPV vaccination was not available yet for. One more limitation is assumption of the rates of having experienced sexual intercourse at each age, which are not derived from official data.
In conclusion, the risk of future cervical cancer differs in accordance with the year of birth. This is a problem that must be overcome not only in Japan but in all countries where the HPV inoculation rate is not increasing.
# Methods
The lifetime risk of cervical cancer was estimated using the latest real data on vaccination rate at each age (2010 through 2015) in Sakai City, Osaka, a government-designated city with a population of 839,128 (2012), regarding the situation on vaccinations. In Sakai City, public funding was initiated in October 2010 and 13 to 16-year-olds were vaccinated. Because vaccinations were initiated mid-year, eligible 16-year-olds could receive public funding by the following September. Periodical vaccinations were initiated from 2013 and 12 to 16-year-olds were vaccinated. The statistical data of the HPV vaccination rates were provided from the Sakai City Government to our study team after approval of the Institutional Review Board of Osaka University Hospital.
Calculation to estimate the lifetime risk of cervical cancer was based on the following hypothesis:
1) The rate of having experienced sexual intercourse at the age of 12, 13, 14, 15, 16, and 17 is assumed to be 0%, 1%, 2%, 5%, 15%, and 25%, respectively. The lifetime rate is assumed to be 85%. 2) HPV infection risk is calculated based on the rate of having sexual intercourse without HPV vaccination at every age and over the lifetime. 3) The relative risk of cervical cancer is proportional to the relative risk of HPV infection over the lifetime. 4) Protective effect against HPV infection is provided by a one-time vaccination (because only the data of the first vaccination was provided) 5) Cervical cancer is caused by HPV infection. The rate of cancer caused by HPV type 16/18 infection is 60%, and 40% by other types, not including infections of HPV type 16/18 in combination with other types. 6) Even if inoculated, the incidence rate of cervical cancer caused by types other than HPV type 16/18 does not change. 7) During the analysis period, the population according to each year of birth does not change. 8) If the vaccine does not exist, the incidence rate of cervical cancer does not change over the course of the analysis period. 9) The presence of vaccination and sexual activity are independent of each other, with sexual activity not changing during the analysis period.
[fig] Figure 2: Figure 1 shows the vaccination rate and the resulting change in cervical cancer relative risk that could actually occur in the future and was already being determined, confirming that our previous prediction is now becoming a reality and severity. [/fig]
|
Inverse Thermoreversible Mechanical Stiffening and Birefringence in a Methylcellulose/Cellulose Nanocrystal Hydrogel
We show that composite hydrogels comprising methyl cellulose (MC) and cellulose nanocrystal (CNC) colloidal rods display a reversible and enhanced rheological storage modulus and optical birefringence upon heating, i.e., inverse thermoreversibility. Dynamic rheology, quantitative polarized optical microscopy, isothermal titration calorimetry (ITC), circular dichroism (CD), and scanning and transmission electron microscopy (SEM and TEM) were used for characterization. The concentration of CNCs in aqueous media was varied up to 3.5 wt % (i.e, keeping the concentration below the critical aq concentration) while maintaining the MC aq concentration at 1.0 wt %. At 20°C, MC/CNC underwent gelation upon passing the CNC concentration of 1.5 wt %. At this point, the storage modulus (G′) reached a plateau, and the birefringence underwent a stepwise increase, thus suggesting a percolative phenomenon. The storage modulus (G′) of the composite gels was an order of magnitude higher at 60°C compared to that at 20°C. ITC results suggested that, at 60°C, the CNC rods were entropically driven to interact with MC chains, which according to recent studies collapse at this temperature into ring-like, colloidal-scale persistent fibrils with hollow cross-sections. Consequently, the tendency of the MC to form more persistent aggregates promotes the interactions between the CNC chiral aggregates towards enhanced storage modulus and birefringence. At room temperature, ITC shows enthalpic binding between CNCs and MC with the latter comprising aqueous, molecularly dispersed polymer chains that lead to looser and less birefringent material. TEM, SEM, and CD indicate CNC chiral fragments within a MC/CNC composite gel. Thus, MC/CNC hybrid networks offer materials with tunable rheological properties and access to liquid crystalline properties at low CNC concentrations.
# Introduction
Cellulose nanocrystals (CNCs) are rodlike, sustainable nanoparticles that can be extracted from wood and plant-based materials by strong acid hydrolysis. [bib_ref] Cellulose Nanocrystals: Chemistry, Self-Assembly, and Applications, Habibi [/bib_ref] Because of the negative surface charges caused by sulfate half-ester residues from the sulfuric acid hydrolysis [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] , CNCs form stable aqueous colloidal suspensions that are isotropic at low concentrations. However, because of their high aspect ratio (typically 10−50), lyotropic liquid crystalline (LC) order emerges when a critical concentration is exceeded, ≥4−5 wt % in the case of cottonbased CNCs. [bib_ref] Helicoidal Self-Ordering of Cellulose Microfibrils in Aqueous Suspension, Revol [/bib_ref] [bib_ref] Rheology and Phase Behavior of Lyotropic Cellulose Nanocrystal Suspensions, Urenã-Benavides [/bib_ref] [bib_ref] Equilibrium Liquid Crystal Phase Diagrams and Detection of Kinetic Arrest in, Honorato-Rios [/bib_ref] The LC assembly is left-handedly twisted and driven by the stacking of the inherently right-handed CNCs and can be detected by optical birefringence via polarized optical microscopy (POM) and optical probes. The chiral nematic pitch and critical concentration can be tuned, for example by adjusting the electrostatic environment of the CNC or through polymer grafting. The cholesteric structure is retained in dried CNC films, which broadens the scope for possible CNCbased photonic, plasmonic, and composite material applications. Because of their excellent mechanical properties, aspect ratio, low density, tunable surface chemistry, and biocompatibility, CNCs have been studied as fillers and reinforcing additives in a variety of compositions ranging from cement paste to composite fibers and hydrogels. Additionally, pristine CNCs form gels at sufficiently high concentrations or when suitably modi-fied. [bib_ref] Rheology and Phase Behavior of Lyotropic Cellulose Nanocrystal Suspensions, Urenã-Benavides [/bib_ref] [bib_ref] Equilibrium Liquid Crystal Phase Diagrams and Detection of Kinetic Arrest in, Honorato-Rios [/bib_ref] [bib_ref] pH-Responsive Cellulose Nanocrystal Gels and Nanocomposites, Way [/bib_ref] However, for their effect to be maximized, it is important to control their assembly over several length scales. [bib_ref] A Sinusoidally Architected Helicoidal Biocomposite, Yaraghi [/bib_ref] Therein, both the LC assembly of CNCs and the interactions with the surrounding matrix are crucial. The packing of CNCs has been studied in detail in confined spaces by capturing CNC suspensions in microdroplets. By tuning the solvent concentrations and evaporation, the LC morphologies within the dried spheres have been effectively controlled. [bib_ref] Confined Self-Assembly of Cellulose Nanocrystals in a Shrinking Droplet, Jativa [/bib_ref] [bib_ref] Hierarchical Self-Assembly of Cellulose Nanocrystals in a Confined Geometry, Parker [/bib_ref] Additionally, by infusing nanoparticles into the droplets, plasmonic, fluorescent, and magnetic properties have been achieved, which has given further insight into the interactions within the cholesteric CNC phase. [bib_ref] Colloidal Cholesteric Liquid Crystal in Spherical Confinement, Li [/bib_ref] On the other hand, methylcellulose (MC, [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] has attracted considerable interest due to its biocompatibility and ability to form thermosensitive gels. [bib_ref] Melting Temperatures of Thermally Reversible Gels IV. Methyl Cellulose-Water Gels, Kato [/bib_ref] It has been shown that, upon heating, the MC chains in aqueous solutions aggregate into persistent fibrils with a hollow, ring-like lateral structure of ∼14 nm diameter and length of several hundreds of nanometers. In water at room temperature (20−22°C), MC chains exist in the form of random coils [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref]. [bib_ref] A Systematic Coarse-Grained Model for Methylcellulose Polymers: Spontaneous Ring Formation at Elevated..., Huang [/bib_ref] Recently, applications combining the properties of CNCs and MC have been pursued, encouraged by the biocompatibility of both components and that their composite gels allow tunable materials. In particular, the aqueous dispersion of methyl cellulose displays inverse thermoreversible properties, ii.e., the free-flowing aqueous dispersion of MC (at room temperature) undergo gelation upon heating (above 40°C) in a reversible manner (sol ↔ gel transition). Inverse thermoreversibility has been observed also in other systems. [bib_ref] Hydrogel Nanoparticle Dispersions with Inverse Thermoreversible Gelation, Hu [/bib_ref] The MC/C hybrids form composite hydrogels that, upon heating from room temperature to 60°C, display a reversibly increased storage modulus. [bib_ref] Thermoresponsive Nanocellulose Hydrogels with Tunable Mechanical Properties, Mckee [/bib_ref] They also form MC/CNC nanocomposite aerogelsand latexes with a double morphology. [bib_ref] Cellulose Nanocrystals and Methyl Cellulose as Costabilizers for Nanocomposite Latexes with Double..., Kedzior [/bib_ref] Several hypotheses have been proposed in the literature, suggesting that MC fibers colloidally wrap the CNCs. [bib_ref] Cellulose Nanocrystals and Methyl Cellulose as Costabilizers for Nanocomposite Latexes with Double..., Kedzior [/bib_ref] [bib_ref] Bionanocomposite Films from Resilin-CBD Bound to Cellulose Nanocrystals, Rivkin [/bib_ref] However, more detailed investigations of the complex interactions and assemblies formed by MC/CNC hydrogel systems have remained unexplored.
In this work, a systematic investigation of mechanical, optical, and morphological details of MC/CNC hydrogels is carried out using multiple complementary techniques. The interplay of rheological and optical (birefringence) properties are studied at room temperature and upon reversible heating from 20°C to 60°C by keeping a fixed aq MC concentration at 1.0 wt % and varying the aq CNC concentration up to 3.5 wt %. The incorporation of CNC rods in a MC hydrogel above a certain concentration, and especially upon heating to 60°C, would modify the rheological properties and optical birefringence due to the inverse thermoreversible nature of MC. The interactions between MC and CNC are explored at room and higher temperatures via isothermal titration calorimetry (ITC). The combination of rheology, quantitative birefringence studies, electron microscopy, and ITC offers new insights into the complex behavior of the MC/CNC hydrogels.
## Experimental section
2.1. Materials. Methylcelluloses (MCs) with different molecular weights of 14,000 g/mol (product no. M7140, lot# SLBQ9046V), 41,000 g/mol (product no. M0262, lot# SLBR8963V), and 88,000 g/ mol (product no. M0512, lot# 079K0054V), hereafter abbreviated as MC14, MC41, and MC88, respectively, were purchased from Sigma-Aldrich. All MC polymers had a methoxy substitution between 27.5 and 31.5% (weight) and a degree of substitution of 1.5−1.9 to produce maximum water solubility as reported by the supplier. Whatman grade 1 qualitative filter papers (cat no. 1001 125), Whatman grade 541 hardened ashless filter papers (cat no. 1541-125, lot# 9722517), and Spectra/Por 1 standard regenerated cellulose dialysis tubing with molecular weight cutoff of 6−8 kDa (part no. 132665, lot# 9200679, Spectrum Laboratories, Inc.) were purchased from VWR. Sulfuric acid (95−97%, product no. 1.00731, lot# K47798131) was bought from Sigma-Aldrich and used as received. Ultrapure Milli-Q water was used in all experiments.
2.2. Cellulose Nanocrystals (CNCs). CNCs were prepared from Whatman grade 1 filter paper following the procedure of Edgar and Gray. [bib_ref] Smooth Model Cellulose I Surfaces from Nanocrystal Suspensions, Edgar [/bib_ref] The filter paper sheets were mechanically ground with a Wiley Mini-Mill (Thomas Scientific, USA) equipped with size 30 filtering mesh to produce homogeneous powder. The resulting powder was hydrolyzed with 64% sulfuric acid. Typically, 272.3 g of sulfuric acid (95−97%) was slowly mixed into 136.2 g of MQ H 2 O to produce the desired concentration. The acid solution was allowed to cool to room temperature before proceeding. Fifteen grams of paper powder was weighed into a 500 mL round-bottomed flask followed by the addition of 64% sulfuric acid solution. The mixture was manually agitated with a glass rod until the paper powder was wetted and submerged into the acid solution. The flask was set into a 45°C water bath and equipped with a Teflon stirring arm, and the paper powder was hydrolyzed under gentle mechanical stirring (32 rpm) for 45 min at 45°C. The reaction was stopped by pouring the reaction mixture into 3 L of MQ H 2 O, i.e., by diluting it ∼10-fold, and it was then left to sedimentate for 20 h. The clear supernatant was discarded by decanting, and the remaining cellulose suspension was washed twice by centrifugation in the following sequence: first, at 6000 rpm for 20 min (Wifug X-3 centrifuge with a fixed angle rotor) and then at 2500 rpm for 45 min, both under ambient conditions. After each centrifugation, the supernatant was discarded by decanting, and the pellet was redispersed in MQ H 2 O. A glass rod was used to homogenize the larger aggregates. The CNC dispersion was further purified by dialysis against MQ H 2 O until the conductivity of the dialysate stayed below 5 μS/cm. Finally, the CNC dispersion was filtered through Whatman 541 filter paper and stored at +4°C until use. Stock solutions containing a 3.5 wt % (35 mg/mL) CNC dispersion were prepared through slow evaporation of water by placing the dispersion in a beaker on a heated magnetic stirrer plate at 250 rpm and 45°C. The solid content of the CNC dispersion was determined gravimetrically by pipetting 1 mL of the dispersion on a watch glass of known weight and placing the watch glass in an oven at 70°C for 20 h to evaporate water and then reweighing the watch glass.
2.3. MC/CNC Composite Hydrogels. MC−CNC gels were prepared following a previously reported procedure. [bib_ref] Thermoresponsive Nanocellulose Hydrogels with Tunable Mechanical Properties, Mckee [/bib_ref] CNC stock dispersion (3.5 wt %) was first diluted to the desired concentrations ranging from 0.5 to 3.5 wt % using MQ H 2 O in volumes close to 10 mL in 40 mL glass vials. An appropriate amount of dry MC powder was then added to the CNC dispersion to obtain the desired concentration. The mixture was first manually agitated to ensure the dispersion and wetting of MC and then further stirred and dissolved by magnetic stirring (250 rpm) for 48 h under ambient conditions. The as-prepared gels were stored in a refrigerator at 4°C (at least 20 h) until use. DLS and zeta (ζ) potential measurements were performed with a Zetasizer Nano ZS90 (Malvern Instruments). Twelve mm square polystyrene cuvettes (product no. DTS0012, Malvern Instruments) were used for DLS, and folded capillary zeta cell cuvettes (product no. DTS1070, Malvern Instruments) were used for zeta potential measurements. Samples of MC88 0.25 wt %, CNC 0.5 wt %, and MC/CNC mixtures with the following compositions 0.25/0.25, 0.25/ 0.5, and 0.5/0.5 (wt %/wt %) were used for ζ-potential measurements. All samples contained 1.0 mM of NaCl. For hydrodynamic size distribution, more dilute samples of MC 0.1 wt %, CNC 0.2 wt %, and MC/CNC 0.1 wt %/0.1 wt % were used. The samples were prepared as described above for MC/CNC hydrogels. The reported distributions and values are the average of three measurements.
2.5. Conductometric Titration. The sulfate group content of CNCs was determined using a conductometric titrator 751 GPD Titrino (Methrom AG) together with Tiamo software as described in the literature. [bib_ref] Cellulose Nanocrystal-Mediated Synthesis of Silver Nanoparticles: Role of Sulfate Groups in Nucleation..., Lokanathan [/bib_ref] Before the titration, the sulfate groups on CNCs were protonated by adding concentrated hydrochloric acid (HCl) to the CNC dispersion so that the final HCl concentration was 0.1 M. The mixture was incubated for 15 min after which the excess acid was removed by dialysis against Milli-Q water until the conductivity of the dialysate remained below 5 μS/cm. For the titration, 20 mL of the obtained CNC dispersion was added to 490 mL of degassed Milli-Q water followed by addition of 0.5 mL of 0.1 M HCl and 1.0 mL of 0.5 M NaCl. The resulting dispersion was titrated against 0.1 M NaOH under constant stirring at 300 rpm. Sodium hydroxide (NaOH) solution was added in 0.02 mL increments every 30 s. The acidic sulfate content was calculated as described in the standard procedure (SCAN-CM 65:02) as a ratio of the NaOH (in μmol) required to neutralize the sulfate groups to the amount of CNC (g) in the dispersion.
2.6. Rheological Measurements. Rheological properties of the gels were determined following the procedures reported previously. [bib_ref] Thermoresponsive Nanocellulose Hydrogels with Tunable Mechanical Properties, Mckee [/bib_ref] TA AR2000 stress-controlled rheometer with a 20 mm steel plate−plate geometry and a Peltier heating system was used. Systematic measurements at different CNC concentrations, ranging from 0 to 3.5 wt %, at a fixed concentration of 1.0 wt % of MC88 were performed at 20 or 60°C. Additionally, samples prepared with 0.5 wt % MC88 concentration and with 1 wt % of MW14 or MC41 combined with 3 wt % of CNCs were measured as reference. Samples were thermally equilibrated for 1 min prior to measurements. Strain-dependent rheological properties were determined at an oscillation frequency of 6.283 rad/s, and frequency sweeps were done at 1% strain at both 20 and 60°C. The samples were covered with a sealing lid to curtail evaporation during the measurement. For the characterization of cyclic temperature response, the samples were heated from 20 to 60°C and cooled to 20°C with an increment of 3.0°C/min. In the temperature sweeps, the storage modulus (G′) and loss modulus (G″) were recorded at an angular frequency of 6.283 rad/s and 1.0% strain.
2.7. Polarized Optical Microscopy (POM). Polarized optical microscopy was performed with Leica DM4500 P high-end polarization microscope combined with a Canon EOS 70D DSLR camera. For studying changes in the temperature-dependent birefringence, MC88/ CNC gels were imaged during a 20−70−20°C temperature cycle with 10°C increments. The heating rate was of 3°C/min, and the samples were allowed to thermally equilibrate for 30 s before acquiring each image. A Linkam LTS 350 heating stage together with TMS 94 temperature programmer and LNP liquid nitrogen pump (Linkam Scientific Instruments, United Kingdom) were used for temperature control. Fixed, identical microscope and camera settings were used for all samples to allow the quantification and comparison of the birefringence intensities. The constant settings were first determined based on the sample with the highest concentration (MC88/CNC 1.0 wt %/3.5 wt %), which presented the maximum birefringence to avoid saturation of the images. The birefringence intensity was then determined as the image mean intensity value using ImageJ. [bib_ref] An Open-Source Platform for Biological-Image Analysis, Schindelin [/bib_ref] [bib_ref] NIH Image to ImageJ: 25 Years of Image Analysis, Schneider [/bib_ref] The gel samples were prepared by placing the sample material inside a SecureSeal imaging spacer (Grace Bio-Laboratories inc., USA) attached on a microscope glass slide, which was then covered with a cover glass. Samples with MC14 and MC41 were imaged under similar conditions for reference purposes. For acquiring colored images, in addition to crossed polarizers, a full-wave plate was used for enhanced colors.
2.8. Transmission Electron Microscopy (TEM). TEM imaging of CNCs and MC88/CNC gels was performed with a JEM-2800 (JEOL) high-resolution TEM microscope operating at 200 kV. Gel samples were prepared on plasma cleaned (30 s, Gatan Solarus 950) and preheated (65°C) C-flat 200 mesh copper grids with holey carbon support film (Electron Microscopy Sciences). Dilute MC/CNC dispersions (MC88/ CNC 0.3 wt %/0.0 and 0.3 wt %/0.3 wt %) were prepared and heated to 65°C for 15 min. Three microliters of gel was then pipetted onto the TEM grid, gently blotted with filter paper edge, and allowed to dry completely at 65°C. For samples prepared at room temperature, 3 μL of gel was then pipetted onto the TEM grid, gently blotted with filter paper edge, and allowed to dry under ambient conditions for 24 h. Plasmacleaned 300 mesh copper grids with carbon-only support film (Electron Microscopy Sciences) were used for the CNC samples. Ten microliters of dilute CNC solution was pipetted onto the grid, incubated for 1 min at RT, and then blotted with filter paper.
2.9. Scanning Electron Microscopy (SEM). SEM imaging was performed with a Zeiss Sigma VP scanning electron microscope at 1− 1.5 kV acceleration voltage. For sample preparation, a droplet of ∼10 μL of MC88/CNC mixture was pipetted on an aluminum SEM stub with carbon tape. The sample was frozen by plunging it in liquid propane for 45 s and then transferring it into liquid nitrogen for at least 5 min and then freeze-dried in a vacuum overnight to obtain aerogels. Alternatively, 200 μL of sample material was directly pipetted into a 1.5 mL microcentrifuge tube and frozen by immersing in liquid nitrogen for 5 min and then freeze-dried in a lyophilizer (0.016 mbar, −100°C) for 20 h. The aerogel samples were attached on aluminum SEM stubs with carbon tape and coated with 5 nm of gold−palladium by using a Leica EM ACE600 high vacuum sputter coater prior to imaging. For preparation of the sample solutions, CNCs were dialyzed against MQ H 2 O until the conductivity of the dialysate was below 5 μS/cm (pH 5.24). The excess dialysis water was then used to prepare the desired MC solutions and to dilute CNC accordingly. All solutions were degassed before the measurements, which were performed at both 25 and 60°C under constant stirring (307 rpm). At 25°C, the elapsed time between the successive injections was 360 s and 600 s at 60°C. Titration heat signals were processed with MicroCal Origin data analysis software. For curve fitting, one set of sites model was applied, and the CNC was assumed to behave as a ligand. The heats of dilution of adding CNC into the dialysis water without MC, at both 25 and at 60°C, were used as reference and subtracted from the respective measurement runs (see Supporting Information, [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref].
# Results and discussion
The morphology, surface charge, and aggregation behavior of freshly prepared CNCs dispersed in aqueous media were characterized with TEM, DLS, and POM. The mean length and width of the CNCs were 181 ± 69 and 7 nm, respectively, as determined from TEM image analysis (seefor size distribution analysis), typical for cotton-based CNCs. [bib_ref] Cellulose Nanocrystal-Based Materials: From Liquid Crystal Self-Assembly and Glass Formation to Multifunctional..., Lagerwall [/bib_ref] The CNC ζ-potential was −64 mV, implying Biomacromolecules Article excellent colloidal stability due to Coulombic interactions, and the sulfate content was determined by conductometric titration to be 234 μmol/g (see Supporting Information, [fig_ref] Figure 4: Scanning electron microscopy images of freeze-dried aerogel samples [/fig_ref] , which is in agreement with previous reports. [bib_ref] Cellulose Nanocrystals and Methyl Cellulose as Costabilizers for Nanocomposite Latexes with Double..., Kedzior [/bib_ref] Polarizing optical microscope (POM) studies of the CNC dispersions showed birefringence at high concentrations, indicating liquid crystalline packing (seefor POM images of CNCs at different concentrations), a typical property of sulfuric acid-hydrolyzed CNCs. [bib_ref] Helicoidal Self-Ordering of Cellulose Microfibrils in Aqueous Suspension, Revol [/bib_ref] [bib_ref] Rheology and Phase Behavior of Lyotropic Cellulose Nanocrystal Suspensions, Urenã-Benavides [/bib_ref] [bib_ref] Chiral Nematic Suspensions of Cellulose Crystallites; Phase Separation and Magnetic Field Orientation, Revol [/bib_ref] As qualitatively shown in [fig_ref] Figure 6: ITC titration curves and respective fits [/fig_ref] for pure CNC). The strain sweeps revealed a linear viscoelastic region up to ∼10% strain at 20°C [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref]. Beyond that, a declining storage modulus G′ was observed, indicating a slight shear thinning. Ideal viscous fluids are expected to show the storage and loss moduli to scale with the angular frequency as G′ ∼ ω 2 , G″ ∼ ω 1 at small ω with G′≪ G″, whereas ideal elastic gels are characterized by G′ ∼ ω 0 , G″ ∼ ω 0 , and G′ ≫ G″.Pure MC88 (1.0 wt %, [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref] shows G′ ∼ ω 1.7 , G″ ∼ ω 0.9 , and G′ ≪ G″ at low ω, i.e., viscous fluid behavior, even though not fully ideal. At 20°C, upon adding CNC, the moduli increased, and their scaling behavior began to indicate a gel-like response. At 60°C, the rheological behavior was different. Even pure MC (aq 1.0 wt %) displayed gel-like response with G′ = 10 Pa and G″ = 2 Pa. In all MC88/CNC mixtures at 60°C, G′ was clearly higher than G″, both independent of ω, characteristic for elastic gels. The samples also turned turbid upon heating. At 60°C, the low strain limit G′ values increased from 12 Pa (for MC88/CNC 1.0 wt %/0.0 wt %) to 540 Pa (for MC88/ CNC 1.0 wt %/3.5 wt %), i.e., a significant increase. At 60°C, strain sweeps of the hybrid gels typically showed strain hardening above 30% strain [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref]. In general, the elastic moduli were reversibly increased by an order of magnitude due to heating, as depicted in. By adjusting both the CNC concentration and temperature, the storage modulus was in total adjustable from 1.6 Pa (for MC88/CNC 1.0 wt %/0.0 wt % at 20°C) to 850 Pa (for MC88/CNC 1.0 wt %/3.5 wt % at 60°C), i.e., over 530fold increase. Hysteresis was observed in consecutive heating and cooling scans, though the moduli values consistently returned to close to the original values .illustrates that the storage moduli increased steeply until the CNC concentration reached ∼1.5 wt % (with MC88 concentration fixed at 1.0 wt %), beyond which a plateau and clear gelation was obtained. This suggests a mechanical percolation phenomenon to be discussed later in the context of birefringence. Altogether, the findings are consistent with the previous rheological studies, [bib_ref] Thermoresponsive Nanocellulose Hydrogels with Tunable Mechanical Properties, Mckee [/bib_ref] and the percolation, not thoroughly recognized thus far, was found relevant to our mechanistic observations. Similar trends were observed for the hybrid gels based on lower molecular weight polymers MC14 and MC41 even though limited to lower moduli [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref].
3.2. Morphology. MC88/CNC hydrogels were imaged by TEM. The pure colloidal CNCs were clearly resolved in TEM [fig_ref] Figure 3: Transmission electron microscopy images [/fig_ref] ; however, resolving the MC component was challenging. Upon heating to 60°C, MC aggregates were expected to form fibrillar structures consisting of hollow ring-like structures of 14 nm lateral sizes and lengths of several hundreds of nanometers [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref]. 47−52 Panels b and c in [fig_ref] Figure 3: Transmission electron microscopy images [/fig_ref] show the TEM micrographs of the MC/CNC 0.3 wt %/ 0.3 wt % sample from 20 and 60°C, respectively. Low concentration was used to better resolve the components. It shows that the CNC was roughly homogeneously distributed within the MC matrix, even though slight CNC bundling cannot be excluded. We were unable to visualize the MC88 fibrils using TEM. Importantly, no CNC alignment was observed in TEM micrographs excluding nematic CNC order.
Freeze-drying allowed another indirect way to investigate the morphologies. [fig_ref] Figure 4: Scanning electron microscopy images of freeze-dried aerogel samples [/fig_ref] shows some of the representative SEM images of aq MC88, CNC, and MC/CNC gels freeze-dried from 20 or 70°C. When aq MC88 (1.0 wt %) was freeze-dried from 20°C using liquid nitrogen, SEM images showed mostly film-like structures with a few fibrillar structures [fig_ref] Figure 4: Scanning electron microscopy images of freeze-dried aerogel samples [/fig_ref]. In contrast, rapid freezing of MC from 70°C showed a pronounced tendency for the fibrillar structures [fig_ref] Figure 4: Scanning electron microscopy images of freeze-dried aerogel samples [/fig_ref] in addition to formation of sheets. This is not unexpected taken the suggested formation of ring-like hollow persistent fibrillar MC aggregates. [bib_ref] A Systematic Coarse-Grained Model for Methylcellulose Polymers: Spontaneous Ring Formation at Elevated..., Huang [/bib_ref] Pure CNC also showed sheets upon freeze-drying [fig_ref] Figure 4: Scanning electron microscopy images of freeze-dried aerogel samples [/fig_ref] with little fibrillar interconnections. On the other hand, when MC/CNC 1.0 wt %/3.5 wt % was freeze-dried from 20°C, sheet-like The sheets were interconnected with MC fibers. The situation further changed for MC/CNC 1.0 wt %/3.5 wt % freeze-dried from 70°C, which showed well-aligned sheets compared to that at room temperature. The high-resolution SEM images clearly showed that the sheets were composed of CNCs, and few interconnecting fibrils were observed. Important for the subsequent conclusions, high-magnification images of sheets (freeze-dried both 20 and 70°C) did not suggest any CNC alignments. This indirectly points toward excluding nematic order in MC/CNC in explaining the birefringence. Instead, CNC chiral fragments within the MC gel network would be the natural source of the birefringence to be discussed next.
3.3. Birefringence. POM was used to explore the birefringence and related liquid crystallinity within the MC88/ CNC gels. Pristine CNC forms cholesteric liquid crystals in water at sufficiently high concentrations (>4−5 wt %). In the present MC88/CNC gels involving low concentrations (<3.5 wt % of CNC, i.e., less than the critical LC concentration for pure aq CNC), liquid crystallinity would not be expected. Therefore, it was not trivial to observe birefringence in the MC/CNC gels at low CNC concentrations, seefor MC/CNC 1.0 wt %/2.5 wt %. As a reference, aq MC88 1.0 wt % showed no clear birefringence. At 20 and 70°C, clear birefringence was observed in MC88/CNC bicomponent gels for the aq CNC concentrations in the range 1.0−3.5 wt % for the fixed aq MC concentration 1.0 wt %. A surprisingly clear stepwise increase was seen between the CNC concentrations 1.0 and 1.5 wt % at both 20 and 70°C. However, already for the added CNC concentration of 1.0 wt %, slight birefringence emerged [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] , and at a CNC concentration as low as 0.5 wt %, transient shear-induced birefringence could be detected in MC/CNC hybrid gels, which is an order of magnitude lower concentration compared to what is typical for CNCs alone to exhibit birefringence. The birefringence intensity increased with increasing CNC concentration, and the intensity of MC88/ CNC 1.0 wt %/3.5 wt % gel was nearly 2 orders of magnitude higher when compared to the 1.0 wt %/0.5 wt % MC88/CNC gel, for example.
Interestingly, the birefringence critical concentration 1.0−1.5 wt % CNC would also agree with the observed percolation threshold for the storage modulus near 1.5 wt % of CNC for a fixed 1 wt % MC. Putting this in perspective, CNCinduced order within gels has been observed also in alginate− CNC mixtures, [bib_ref] Effect of Jet Stretch and Particle Load on Cellulose Nanocrystal-Alginate Nanocomposite Fibers, Urenã-Benavides [/bib_ref] and in a few cases, birefringence has been reported in cellulose nanofiber gels. [bib_ref] Smalyukh, I. I. Liquid Crystalline Cellulose-Based Nematogels, Liu [/bib_ref] It would be natural to expect that the birefringence, i.e., the optical anisotropy, would be reduced upon heating. By contrast, we observed that the birefringence increased upon heating from room temperature to 70°C and reversibly decreased upon cooling back to room temperature (as qualitatively shown in [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref]. More quantitatively, the extents of birefringence at different compositions and at 20 and 70°C are depicted inand [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref]. The thermoreversible birefringence was reduced for the lower CNC concentrations. Therefore, like the
## Biomacromolecules
Article rheological properties, birefringence showed inverse thermoreversibility.
Regarding the source of the birefringence, TEM and SEM allow for excluding nematic CNC order. Additional insight was gained from CD spectroscopy of pure CNC, MC, and MC/CNC systems.
The pure CNC (3.5%) showed CD signal at 20°C, suggesting that, above a certain concentration (i.e., critical CNC concentration), the CNCs exist as chiral aggregates in an aqueous dispersion [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref]. Heating the aq CNC to 60°C showed no significant change in the CD signal. Interestingly, CD signal was also observed in pure 1.0 wt % MC at 60°C [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] , possibly related to the existence of ring-like fibrillar structures. [bib_ref] Aggregation and Network Formation of Aqueous Methylcellulose and Hydroxypropylmethylcellulose Solutions, Bodvik [/bib_ref] [bib_ref] Anisotropic Self-Assembly and Gelation in Aqueous Methylcellulose-Theory and Modeling, Ginzburg [/bib_ref] [bib_ref] A Systematic Coarse-Grained Model for Methylcellulose Polymers: Spontaneous Ring Formation at Elevated..., Huang [/bib_ref] Remarkably, CD signal was detected from MC/CNC gels at as low as 0.5 wt % of CNC in 1.0% MC. Further, the signals of MC/CNC were amplified by increasing CNC weight fractions as well as upon increasing the temperature. A temperature-dependent peak amplification and shift suggested possible structural deformations and the thermoresponsive nature of the CNC aggregates [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref].
3.4. MC−CNC Interactions. In order to shed light on the inverse thermoreversible gelation and optical anisotropy, ITC was performed using MC/CNC 0.4 wt %/0.45 wt %, i.e., dilute concentrations, at 25°C and upon heating to 60°C, using MC14, MC41, and MC88. The titration curves and corresponding fits are shown in [fig_ref] Figure 6: ITC titration curves and respective fits [/fig_ref] and [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] for mixtures with different MC molecular weights. At 25°C, the enthalpy change (ΔH) for the interaction of MC41 and CNC was negative (−3390 kcal/ mol), whereas the entropic contribution (−TΔS) was almost the opposite (3369 kcal/mol), leading to a favorable net negative Gibbs free energy (ΔG = ΔH − TΔS) of −21 kcal/mol. For the other two MCs of different molecular weights, the findings were essentially similar at 25°C [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] , and the overall Gibbs free energy remained negative (−17 kcal/mol and −12 kcal/mol for MC88 and MC14, respectively), indicating attraction. Therefore, ITC suggests that CNC and MC spontaneously interact at room temperature, driven by enthalpic interactions tentatively based on hydrogen bonding and van der Waals interactions.
The situation drastically changes at 60°C [fig_ref] Figure 6: ITC titration curves and respective fits [/fig_ref] and Figures S15). At this temperature, the enthalpy change was strongly positive (22630 kcal/mol for MC41), presumably highlighting the endothermic nature of the heat-induced MC gelation. The entropy factor (−TΔS) became strongly negative (−22654 kcal/mol), leading to a net negative Gibbs free energy change of −24 kcal/mol in the case of MC41/CNC. Therefore, at 60°C, entropy was the driving force, but regardless of the change of the interaction of the components, the process remained spontaneous. Unfortunately, with the MC of the highest molecular weight MC88, the increased viscosity at 60°C interfered with the measurements, making them less reliable. They gave only one reasonably coherent data set that produced a clearly deviating positive net ΔG of 67 kcal/mol. Nonetheless, the entropy-driven behavior appeared characteristic for all MC/ CNC mixtures. Therefore, the roles of enthalpy and entropy completely overturned upon increasing the temperature. Positive enthalpic contribution due to gelation opposed the interaction, whereas entropic gain suggested hydrophobic interactions between MC and CNC and entropically favorable consequent structural adjustments of the fibrous MC, such as partial or local disassembly or release of water from the fibrils' solvation layer. This potentially contributes to the lack of visible fibrillar MC structures (TEM in. We also point out that all other samples, except those used in the ITC experiments, were prepared at room temperature, and the behavior at 60°C was explored after the heating. Therefore, they were expected to involve history dependence upon heating, and the thermoreversible behavior can be more complex than suggested in the ITC experiments, where the samples were directly prepared at 60°C.
3.5. Rationalization of the Results. Next, we suggest an explanation by combining our experimental observations and previous literature reports, see the schematics in [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref]. First, as shown in [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] , MC chains at 20°C exist in random coiled conformations. The chains are expected to be highly entangled as the concentration used in our experiments (1.0 wt % MC) is much above the overlap concentrations reported for methyl cellulose and other cellulose derivatives. [bib_ref] Comparative Rheological Behavior of Some Cellulosic Ether Derivatives, El Ghzaoui [/bib_ref] Further, it has been shown that the persistence length (L p ) for MC solutions at room temperature is ∼12−17 nm. Upon heating, the viscous fluid turns into a gel due to the formation of shape-persistent, long fibrils with ringlike lateral order and diameter of ∼14 nm (see [fig_ref] Figure 1: Components of the composite hydrogel [/fig_ref] , inset) as suggested by the previous models to explain the MC gelation at elevated temperatures [bib_ref] Aggregation and Network Formation of Aqueous Methylcellulose and Hydroxypropylmethylcellulose Solutions, Bodvik [/bib_ref] [bib_ref] Anisotropic Self-Assembly and Gelation in Aqueous Methylcellulose-Theory and Modeling, Ginzburg [/bib_ref] [bib_ref] A Systematic Coarse-Grained Model for Methylcellulose Polymers: Spontaneous Ring Formation at Elevated..., Huang [/bib_ref] and supported by SEM of aerogels [fig_ref] Figure 4: Scanning electron microscopy images of freeze-dried aerogel samples [/fig_ref]. The temperature-dependent stiffening of pure MC gels also suggests a significant increase in L p at higher temperature.In MC/CNC composite hydrogels, the solid MC was added to the aqueous CNC dispersions; therefore, three possibilities can be assumed. First, when the concentration of CNC is below a critical concentration to induce gelation at room temperature, the adsorption of MC chains and
## Biomacromolecules
Article wrapping of CNCs is involved as reported in the literature [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref]. However, because of more random coiling, the softer but entangled MC in this case leads to a smaller storage modulus. At small CNC aq concentrations, the CNCs have minimal mutual steric interactions, leading to reduced reinforcement [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref]. Increasing the concentration above the percolation threshold, i.e, 1.5 wt % of CNCs led to rapid gelation at 20°C, where MC chains interact enthalpically with CNCs as suggested by the ITC experiments. In this case, the CNCs exist as small chiral aggregates that are kinetically trapped within the MC network [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref] , thus inducing some birefringence. At high temperature, MC/CNC gels showed increased storage modulus passing the percolation threshold at aq CNC 1.5 wt % with the fixed aq MC 1.0 wt % and creating a liquid crystalline order of CNC manifested as strong birefringence [fig_ref] Figure 7: Schematics for the MC/CNC composite gels [/fig_ref]. Note, however, that in MC/CNCs composite gels, the behavior of MC may be more complex than in the case of pure MC due to the interplay between CNC and MC. ITC suggests that, at high temperature, CNCs are entropically attracted to MC. SEM, TEM, and CD suggest that the birefringence is not due to nematic order but rather due to the chiral aggregates of CNC. Thus, CNC chiral aggregates can be expected to become topologically entrapped within the persistent fibrillar MC, thus effectively increasing the storage modulus.
# Conclusions
Mixtures of MC and CNC form thermoreversible, tunable, and modular multifunctional bicomponent hydrogels where the liquid crystalline order and rheological properties can be tuned by temperature in a reversible manner. Using a fixed MC concentration of 1 wt % and varying the CNC loading in the range 0−3.5 wt % (below the critical LC concentration of pure aq CNC), and by controlling the temperature (20−70°C), the rheological storage modulus is tunable in a range spanning almost three orders of magnitude. Upon increasing the CNC concentration past 1.0−1.5 wt %, the storage modulus of MC/ CNC hybrids acquired a plateau, and the birefringence increased stepwise, suggesting a percolative phenomenon. Upon heating to 60−70°C followed by recooling back to room temperature, the storage modulus and birefringence reversibly increased. The findings point toward reversible formation of CNC chiral aggregates, where the thermally induced space-filling aggregation within the gel network leads to stiffening. The findings suggest that detailed understanding of thermoreversible colloidal reinforcing filler and matrix interactions can allow unexpected new multiresponsive and functional biocomposites and thermosensitive biomaterials for future applications.
## ■ associated content
## * s supporting information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.biomac.8b00392.
Additional material regarding rheological properties, birefringence, conductometric titration data, dynamic light scattering data, zeta potential measurements, CD spectroscopy data, and ITC titration data (PDF)
## ■ author information
Corresponding Authors *E-mail: [email protected]. *E-mail: [email protected].
[fig] Figure 1: Components of the composite hydrogel: The chemical structures of (a) cellulose nanocrystal (CNC) showing the repeat units with a scheme of the colloidal rod and (b) methyl cellulose (MC). (c) At 20°C, aq MC behaves as individual polymer chains. (d) At 60°C, the aq MC chains are supramolecularly aggregated into shape-persistent fibrillar assemblies with hollow ring-like lateral structure. [/fig]
[fig] Figure 2: (a) Gelation of aq MC88, CNC, and MC88-CNC composites at 20 and 60°C as shown by vial inversion. The compositions are given in aqueous wt %. (b) Storage moduli as determined with dynamic rheology of MC88/CNC hydrogels as a function of temperature (b) for three compositions and as a function of CNC concentration (c) for 20 and 60°C (for comparison, see Figure S9 using the linear G′ scale). Note that in (b) and (c), the aq MC concentration is fixed at 1 wt %, and the aq CNC concentration is varied. Biomacromolecules Article DOI: 10.1021/acs.biomac.8b00392 Biomacromolecules 2018, 19, 2795−2804 structures were observed that were largely composed of CNCs as a close-up shows random CNC structures (Figure 4e,f). [/fig]
[fig] Figure 3: Transmission electron microscopy images. (a) TEM image of CNC 0.02 wt % shows well-dispersed nanocrystals in water at 20°C. (b) TEM image of MC/CNC 0.3 wt %/0.3 wt % hydrogel at 20°C. (c) TEM image of MC/CNC 0.3 wt %/0.3 wt % hydrogel at 60°C. [/fig]
[fig] Figure 4: Scanning electron microscopy images of freeze-dried aerogel samples. (a) MC88 1.0 wt % freeze-dried from 20°C. (b) MC88 1.0 wt % freeze-dried from 70°C. (c,d) CNC 3.5 wt % freeze-dried from 20°C. (e,f) MC/CNC 1.0 wt %/3.5 wt % freeze-dried from 20°C. (g,h) MC/ CNC 1.0 wt %/3.5 wt % freeze-dried from 70°C.Biomacromolecules Article DOI: 10.1021/acs.biomac.8b00392 Biomacromolecules 2018, 19, 2795−2804 [/fig]
[fig] Figure 5: (a) The MC/CNC gels were strongly birefringent as demonstrated by POM imaging of MC88/CNC 1.0 wt %/2.5 wt % at 20°C. (b) Birefringence as a function of CNC concentration by keeping the concentration MC fixed at 1.0 wt %. Note the stepwise percolative increase of birefringence between 1.0 and 1.5 wt % of CNC. Note the logarithmic scale. (c) POM micrographs of MC88/CNC 1.0 wt %/3.5 wt % at room temperature, upon heating to 70°C, and upon recooling to room temperature, showing a reversible increase of the birefringence upon heating. Scale bars, 200 μm. (d) Thermoreversible increase of birefringence for three MC/CNC compositions. Note that a linear scale is selected in this case. [/fig]
[fig] Figure 6: ITC titration curves and respective fits (a) at 25°C and (b) at 60°C for MC41/CNC showing that, at room temperature, the MC and CNC components bind enthalpically. By contrast, at 60°C the entropy dominates, suggesting hydrophobic interactions and structural adjustments, thus promoting CNC packing with the MC gel network. [/fig]
[fig] Figure 7: Schematics for the MC/CNC composite gels. (a) Below the CNC critical concentration at 20°C. (b) Above the CNC critical concentration at 20°C. (c) Above the CNC critical concentration at 60°C when the MC chains have the tendency to aggregate into helically winding fibrils. The suggested entanglements within MCs are highlighted with red circles. [/fig]
|
This review article aims to: (1) discern from the literature the immune and inflammatory processes occurring in the pericardium following injury; and (2) to delve into the molecular mechanisms which may play a role in the progression to constrictive pericarditis. Pericarditis arises as a result of a wide spectrum of pathologies of both infectious and non-infectious aetiology, which lead to various degrees of fibrogenesis. Current understanding of the sequence of molecular events leading to pathological manifestations of constrictive pericarditis is poor. The identification of key mechanisms and pathways common to most fibrotic events in the pericardium can aid in the design and development of novel interventions for the prevention and management of constriction. We have identified through this review various cellular events and signalling cascades which are likely to contribute to the pathological fibrotic phenotype. An initial classical pattern of inflammation arises as a result of insult to the pericardium and can exacerbate into an exaggerated or prolonged inflammatory state. Whilst the implication of major drivers of inflammation and fibrosis such as tumour necrosis factor and transforming growth factor β were foreseeable, the identification of pericardial deregulation of other mediators (basic fibroblast growth factor, galectin-3 and the tetrapeptide Ac-SDKP) provides important avenues for further research.Core tip: Constrictive pericarditis arises as a complication of pericarditis from a wide range of aetiologies. A comprehensive understanding of the fibrotic process eventually leading to pathological symptoms is currently lacking.Through this review of the literature, we have identified various molecular mediators which are likely to play a role in the establishment of constriction and which warrant further studies.Ramasamy V, Mayosi BM, Sturrock ED, Ntsekhe M. Established and novel pathophysiological mechanisms of pericardial injury and constrictive pericarditis. World J Cardiol 2018; 10(9): 87-96 Available from: URL
# Introduction
Pericarditis describes the clinical syndrome that occurs in response to injury of the pericardium. Following an episode of pericarditis, the natural history of the disease is variable and unpredictable. In a significant proportion of patients there is progression from acutely inflamed pericardium to chronic thickening, fibrosis, and fusion of the two pericardial layers with often dire consequences for patients. The molecular mechanisms involved in the inflammatory and immune mediated injury during pericarditis and the mechanisms involved in progression to constrictive pericarditis are poorly understood . Such an understanding may be important to the design and development of interventions which are able to interrupt and prevent maladaptive and deleterious pericardial responses.
We conducted a comprehensive review of the available literature to summarize what is currently known about (1) immune and inflammation-mediated pericardial injury in a range of different causes of pericarditis; and (2) the molecular mechanisms involved in both pericarditis and subsequent post inflammatory progression to fibrosis and constrictive pericarditis.
## Search strategy
The literature search was conducted in Pubmed, Embase, ScienceDirect and Google Scholar to identify journal articles for the review that had been published up to July 2017. In order to identify papers describing inflammatory and fibrotic processes occurring in the different types of pericarditis, the following search terms were used: "Pericarditis"and ("inflammation" or "fibrosis" or "constrictive" or "constriction"). The search was repeated for the common types of pericarditis described using the following search terms in the search criteria described above: "(uraemic or uremic)", "tuberculous", "malignant", "radiation", "autoimmune", "viral", "infectious", "post surgery" and "myocardial infarction" [bib_ref] Yacoub MH; Task Force on the Diagnosis and Management of Pricardial Diseases..., Maisch [/bib_ref]. Literature regarding different animal models of pericarditis was obtained using the search terms: "animal model" and "pericarditis". Papers were then filtered according to their titles and content for the identification of relevant literature.
## Normal pericardium
The pericardium is double layered flask-like sac which encloses the heart through its attachments to the great vessels, namely the vena cava, aorta, and pulmonary artery and vein. It is lined on the outside by the parietal pericardium, which is a fibrous layer of connective tissue rich in elastic fibres and collagenous fibres. This fibrosa is supplied by a network of blood and lymphatic vessels that contains macrophages and fibroblasts. The inner visceral pericardium is a single serous layer composed of flat, irregular, ciliated mesothelial cells resting on a thin basement membrane and separated from the fibrous layer by a thin sub-mesothelial space [bib_ref] Histologic and ultrastructural features of normal human parietal pericardium, Ishihara [/bib_ref] [bib_ref] Mesothelial cells: their structure, function and role in serosal repair, Mutsaers [/bib_ref]. These two layers of the pericardium are 1-2 mm thick and give rise to a cavity which contains on average 15 to 35 mL of pericardial fluid, under normal physiological conditions [bib_ref] Pericardial disease, Little [/bib_ref]. Pericardial fluid is formed from ultrafiltration of plasma and comprises largely globular proteins, phospholipids and surfactant-like prostaglandins .
## Pathophysiological response of the pericardium to injury
## Pericarditis
The pathophysiological response of the pericardium to injury is characterized by intense inflammation with or without effusion, and the clinical syndrome of pericarditis. Pericarditis is a common disorder [bib_ref] Evaluation and Treatment of Pericarditis: A Systematic Review, Imazio [/bib_ref] that can result from both an infectious and non-infectious aetiology and presents clinically as pericarditis with and without effusion. Causes of pericarditis include viral, bacterial, fungal, uraemic, post-acute myocardial infarction, neoplastic, post-cardiac surgery, following mediastinal irradiation and as a consequence of systemic autoimmune diseases [bib_ref] Yacoub MH; Task Force on the Diagnosis and Management of Pricardial Diseases..., Maisch [/bib_ref]. The major complications of pericarditis are cardiac tamponade with and without hemodynamic instability in the short term, constrictive pericarditis in the long term and death [bib_ref] Pericardial disease, Little [/bib_ref]. The latter is usually the consequence of chronic inflammation, thickening, adhesion, fibrosis and obliteration of the pericardial space.
## Clinical models of post inflammatory pericarditis
## Uraemic pericarditis
Uraemic pericarditis is a complication of acute and chronic renal failure, which can arise prior to, and on dialysis treatment. The condition was prevalent before the widespread use of dialysis and was commonly associated with a poor prognosis and high mortality.
Currently, with modern dialysis, it has a highly improved prognosis and survival rate [bib_ref] Lessons from the past: Uremic pericarditis, De Gouveia [/bib_ref]. The mechanism of the development of pericarditis in uraemic disease is poorly understood. Although pericarditis is more frequent in cases of severe uraemia, there is no correlation between blood urea and creatinine levels and the appearance of pericarditis [bib_ref] Pericarditis in renal disease, Kumar [/bib_ref]. Uraemic pericarditis is usually exudative with protein and large numbers of mononuclear cells in the pericardial fluid [bib_ref] Subacute constrictive uremic pericarditis, Reyman [/bib_ref]. Serous or haemorrhagic effusions are common and typically evolve into a fibrinous state. This fibrotic state often manifests as a rough granular surface with irregular, scattered adhesions between parietal and visceral pericardium in a "bread and butter" pattern [bib_ref] Lessons from the past: Uremic pericarditis, De Gouveia [/bib_ref] [bib_ref] Chronic constrictive pericarditis following uremic hemopericardium, Lindsay [/bib_ref]. However, densely adherent pericarditis and gross pericardial thickening with organizing fibrinous pericarditis have been found at autopsy in cases of uraemic disease [bib_ref] Uremic pericarditis. Clinical features and management, Bailey [/bib_ref].
## Radiation pericarditis
Radiation pericarditis occurs as a complication of radiation therapy of malignant mediastinal tissues and organs, most commonly breast cancer or mediastinal Hodgkin's disease. Radiation pericarditis was a common adverse outcome when large areas of the heart were exposed to high doses of radiation therapy, but the advent of chemo-and immuno-therapy has decreased the incidence of the condition [bib_ref] Radiation-induced cardiovascular diseases: is the epidemiologic evidence compatible with the radiobiologic data?, Schultz-Hector [/bib_ref]. Both acute and late post radiation pericardial injuries have been described. Acutely, radiation toxicity can cause micro-vascular damage and episodic pericardial ischemia, which in turn leads to permeable neovascularization and fibrous deposition. Later, activated fibroblasts express increasing type Ⅰ collagen levels, with subsequent focal massive hyalinization, adhesions of the epicardium and thickening pericardium [bib_ref] Radiation myocardial fibrosis simulating constrictive pericarditis. A review of the literature and..., Botti [/bib_ref] [bib_ref] Management of patients with radiation-induced pericarditis with effusion: a note on the..., Morton [/bib_ref]. There is also evidence of impaired drainage of extracellular fluid from the pericardium, a chronic fibrinous exudative pericarditis and vascular and lymphatic fibrosis [bib_ref] Radiation-induced heart disease: pathologic abnormalities and putative mechanisms, Taunk [/bib_ref]. The degree of inflammation and thickening in radiation pericarditis corresponds to the X-ray exposure, with marked thickening observed more at the site of irradiation. This suggests a cellular injury and necrosis induced inflammatory response as a result of the acute radiation of actively proliferating cells, potentially mesothelial cells of the pericardium. On a molecular level, increased collagen synthesis and the pathological remodelling of the pericardium post radiation have been associated with the activation of various growth factors and cytokines including transforming growth factor β (TGF-β) and connective tissue growth factor (CTGF) [bib_ref] Pathogenetic mechanisms in radiation fibrosis, Yarnold [/bib_ref].
## Autoimmune disease pericarditis
Pericardial involvement can arise in various autoimmune diseases, most commonly systemic lupus erythematosus (SLE), rheumatoid arthritis (RA) and systemic sclerosis (SSc) [bib_ref] Yacoub MH; Task Force on the Diagnosis and Management of Pricardial Diseases..., Maisch [/bib_ref]. SLE is a chronic inflammatory disease with a broad range of clinical manifestations and a variable disease course. The exact aetiology of SLE is still unclear, but it is likely to be mediated by antibodies and immune complexes (IC) which typically contribute to the clinical manifestation of SLE. Immune complexes can result in complement activation and inflammation and they have been detected in the pericardial fluid in SLE [bib_ref] Immune complexes in the pericardial fluid in systemic lupus erythematosus, Quismorio [/bib_ref]. While pericarditis is the most common cardiac manifestation of SLE, constrictive pericarditis is a rare occurrence [bib_ref] Constrictive pericarditis in systemic lupus erythematosus. Demonstration of immunoglobulins in the pericardium, Jacobson [/bib_ref]. RA is a chronic inflammatory disorder that primarily affects joints. Symptomatic pericarditis arises in less than 10% of patients with severe disease and is often associated with a poor prognosis. The pericardial involvement is usually a diffuse pericardial effusion, sometimes associated with leukocyte infiltration and often positive for rheumatoid factor and immune complexes. Constrictive pericarditis is not common in RA and can arise despite second-line therapy. Thickened pericardia with collagenous fibrous tissue and organising fibrin, fibrinous exudate and leukocyte infiltration have been described [bib_ref] The heart and cardiovascular manifestations in rheumatoid arthritis, Voskuyl [/bib_ref]. Asymptomatic pericardial effusions occur in upto 30%-50% of patients with RA and represents the most common cardiac manifestation of the disease.
Systemic sclerosis is a systemic autoimmune disease characterized by aberrant fibroblast activity resulting in dense fibrosis of visceral organs and skin. Pericardial manifestations include pericardial effusions, fibrous pericarditis, pericardial adhesions or constrictive pericarditis. Clinical manifestations of pericardial pathology are apparent in over 5%-16% of cases. The pathogenesis of pericardial effusions in SSc is believed to differ from the inflammatory pathway triggered by auto-antibodies and immune complexes of SLE and RA as evidenced by the "non-inflammatory" profile of the pericardial fluid. Instead, the release of basic fibroblast growth factor (bFGF) and histamine by mast cells may contribute to the pathophysiological manifestations [bib_ref] Cardiac manifestations in systemic sclerosis, Lambova [/bib_ref] [bib_ref] Pericardial involvement in systemic sclerosis, Byers [/bib_ref].
The inflammatory basis of autoimmune pericarditis, centered around the role of the inflammasome, has recently been reviewed by Xu et al [bib_ref] New Insights into Pericarditis: Mechanisms of Injury and Therapeutic Targets, Xu [/bib_ref].
## Post myocardial infarction pericarditis
Pericarditis is a common sequelae of transmural myocardial infarction (MI) and arises "early" as pericarditis epistenocardica or as a "delayed" presentation in the form of Dressler syndrome [bib_ref] Pericarditis Epistenocardica or Dressler Syndrome? An Autopsy Case, Feola [/bib_ref]. The acute form of pericarditis is often diagnosed 1-4 d post MIand is sometimes accompanied by a pericardial effusion. Vascular injury and myocardial necrosis have been associated with increased incidence of pericarditis, suggesting an inflammatory response to injury [bib_ref] Regional pericarditis: a review of the pericardial manifestations of acute myocardial infarction, Dorfman [/bib_ref] [bib_ref] Factors associated with pericardial effusion in acute Q wave myocardial infarction, Sugiura [/bib_ref]. Fibrous deposits and adhesions often develop in the visceral and parietal pericardium covering the area of infarction but may also involve wider and more diffuse pericardial surfaces [bib_ref] Pericardial heart disease: its morphologic features and its causes, Roberts [/bib_ref]. Dressler syndrome (DS) commonly arises around by Ristić et al [bib_ref] Pericardial cytokines in neoplastic, autoreactive, and viral pericarditis, Ristić [/bib_ref] , whilst TGF-β levels were only elevated in serum. However, conflicting Interferon-γ (IFN-γ) deregulation has been reported by the two studies, with a strongly elevated levels described by Pankuweit et al [bib_ref] Cytokine activation in pericardial fluids in different forms of pericarditis, Pankuweit [/bib_ref] , whilst no differences were reported by Ristić et al [bib_ref] Pericardial cytokines in neoplastic, autoreactive, and viral pericarditis, Ristić [/bib_ref]. This is probably due to the small sample size used in the latter study. Further, Karatolios et al [bib_ref] Cytokines in pericardial effusion of patients with inflammatory pericardial disease, Karatolios [/bib_ref] measured increased pericardial and serum levels of vascular endothelial growth factor (VEGF) in viral pericarditis as well as decreased bFGF levels in the pericardial fluid.
Elevated serum cardiac troponin Ⅰ (cTnⅠ) levels have been observed in viral pericarditis and have been associated with ST-segment elevation, and pericardial effusion. Whilst, this increase is often more pronounced with increased myocardial inflammation, it did not affect the prognosis and the development of tamponade and fibrosis [bib_ref] Cardiac troponin I in acute pericarditis, Imazio [/bib_ref].
## Tuberculous pericarditis
Tuberculous (TB) pericarditis accounts for roughly 4% of cases of acute pericarditis in the developed world. However, in developing countries with a high prevalence of tuberculosis, around 70% of cases of large pericardial effusion are attributable to TB [bib_ref] A modern approach to tuberculous pericarditis, Syed [/bib_ref] [bib_ref] Tuberculous pericarditis, Fowler [/bib_ref]. Further, HIV co-infection has not only increased the number of TB pericarditis cases, but has also changed its clinical manifestations and therapeutic considerations [bib_ref] Tuberculous pericarditis with and without HIV, Ntsekhe [/bib_ref]. The spread of Mycobacterium tuberculosis (MTb) to the pericardium occurs either through retrograde lymphatic spread or through haematogenous spread from primary sites of infection [bib_ref] Tuberculous pericarditis, Mayosi [/bib_ref] [bib_ref] Constrictive pericarditis: clinical and pathophysiologic characteristics, Myers [/bib_ref]. The inflammatory process in TB pericarditis follows a sequence of pathological events. An early fibrinous exudate is formed with leucocytosis, and early granuloma formation as a response to the high mycobacterial abundance, followed by a sero-sanguineous effusion with a predominantly lymphocytic exudate. The effusion gradually recedes whilst the granulomatous architecture is organised to restrict mycobacterial spread. Fibrin, collagen and extracellular matrix (ECM) deposition lead to pericardial thickening and fibrosis [bib_ref] Tuberculous pericarditis, Mayosi [/bib_ref]. Infection of the pericardium with the bacilli elicits an immune response, stimulating lymphocytes to release cytokines which activate macrophages and influence granuloma formation. This initial reaction presents pathologically with polymorphonuclear leucocytosis and granuloma formation [bib_ref] Tuberculosis and the Heart, Mutyaba [/bib_ref]. Marked elevations of IL-10 and IFN-γ accompanied by low levels of bioactive TGF-β levels in tuberculous pericardial fluid suggest a Th-1 mediated delayed type hypersensitivity response to the pathogen [bib_ref] Prevalence, hemodynamics, and cytokine profile of effusiveconstrictive pericarditis in patients with tuberculous..., Ntsekhe [/bib_ref]. Similarly, Reuter et al [bib_ref] Diagnosing tuberculous pericarditis, Reuter [/bib_ref] measured significantly increased IFN-γ levels in the pericardial fluid and observed large numbers of mesothelial cells in tuberculous pericardial aspirates.
A role for complement fixing antimyolemmal antibodies has also been suggested in the development of exudative tuberculous pericarditis through cardiocyte cytolysis [bib_ref] Immune reactions in tuberculous and chronic constrictive pericarditis. Clinical data and diagnostic..., Maisch [/bib_ref]. More recently, it was shown that the tetrapeptide N-acetyl-seryl-aspartyl-lysyl-proline (Ac-two weeks post MI. It is an uncommon presentation since the advent of early reperfusion therapy with thrombolytic therapy and primary percutaneous intervention, and with the widespread use of heparin. DS is presumably a recurrent immune-inflammatory syndrome arising from the release of auto-myocardial antigens from necrosis of myocardial tissues. The formation of immune complexes are believed to trigger a hypersensitivity reaction from molecular mimicry and cross-reactions [bib_ref] Pericarditis Epistenocardica or Dressler Syndrome? An Autopsy Case, Feola [/bib_ref] [bib_ref] Is Dressler syndrome dead?, Bendjelid [/bib_ref]. Indeed, the presence of increased anti-myocardial antibodies following myocardial injury has been previously suggested and supports a possible autoimmune pathogenesis [bib_ref] Immunological Studies in the Post-Cardiotomy Syndrome, Robinson [/bib_ref].
## Post cardiac surgery pericarditis
Pericarditis can present as a midterm or late complication of cardiac surgery. Post surgery pericarditis often bears restrictive haemodynamic characteristics despite an open pericardium and can occur as early as 2 wk following surgery [bib_ref] Constrictive pericarditis: early and late complication of cardiac surgery, Cohen [/bib_ref]. Adhesions and fibrous patches in the pericardium lead to constrictive pericarditis and cause symptoms of dyspnoea, and signs of congestive cardiac failure. Whilst the exact mechanism for the development of pericardial fibrosis following surgery is obscure, the presence of blood in the pericardial cavity may play a role, with failure to drain bloody effusions being a risk factor for the development of fibrosis. Blood in the pericardium may result in irritation of the serosal layer and inflammation [bib_ref] Clinical characteristics of patients with constrictive pericarditis after coronary bypass surgery, Matsuyama [/bib_ref]. However, pericardial fibrosis in the absence of bloody effusions after surgery has also been documented [bib_ref] Constrictive pericarditis after cardiac surgery, Gaudino [/bib_ref].
## Post infectious pericarditis
A range of infectious organisms can affect the pericardium, but the most common causes are viruses (coxsakie viruses, influenza virus and enteric cytopathogenic human orphan virus, among others) and bacteria (Staphylococcus and Streptococcus, Haemophilus, and M. tuberculosis).
Bacterial (purulent) pericarditis which is a life threatening condition is characterised by gross pus in the pericardium or microscopically purulent effusion [bib_ref] Bacterial pericarditis: diagnosis and management, Pankuweit [/bib_ref]. It is an uncommon occurrence in the developed world, due to widespread antibiotic usage. However, tuberculous pericarditis is a leading cause of pericarditis in Sub Saharan Africa and is discussed separately below [bib_ref] Tuberculous pericarditis, Mayosi [/bib_ref]. Viral pericarditis, on the other hand, is a common manifestation and is often self-limiting, in that only a small number of patients develop fibrous complications. Viral antigens lead to an inflammatory response of lymphocytic predominance which often results in effusions. Cytotoxic and T and/or B cell-driven immune-mediated mechanisms of inflammation have been described in different types of viral infections .
Increased levels of IL-6 and IL-8 have also been described in both serum and pericardial fluid in viral pericarditis, with a marked local increase in the pericardial cavity [bib_ref] Pericardial cytokines in neoplastic, autoreactive, and viral pericarditis, Ristić [/bib_ref] [bib_ref] Cytokine activation in pericardial fluids in different forms of pericarditis, Pankuweit [/bib_ref]. An increase in pericardial TNF-α levels has also been measured in the pericardial fluid Coxsackie B viruses are known to cause perimyocarditis, an acute inflammation of the pericardium and the underlying myocardium [bib_ref] Coxsackie B myopericarditis in adults, Smith [/bib_ref]. In Coxsackie B3 induced perimyocarditis in mice, an early onset of myocardial injury and necrosis has been observed, followed by marked pericardial fibrosis [bib_ref] Coxsackie virus B3 perimyocarditis in BALB/c mice: experimental model of chronic perimyocarditis..., Matsumori [/bib_ref]. In this particular study, the sub-epicardial myocardial tissues appeared to mostly contribute to the fibrotic process, with infiltration by macrophages, lymphocytes and polymorphonuclear leukocytes observed in the myocardial layer. However, in Coxsackie B4 pericarditis in mice, fibrotic lesions occurred independently of, or in conjunction with adjacent myocardial lesions. Similar patterns of inflammation were observed in the mesothelial cells with necrosis, cellular infiltration, inflammatory cell infiltration and fibrinous exudate. The inflammatory processes are likely to be as a result of infection of mesothelial cells by the virus, since viral antigens have been detected in the mesothelial cells [bib_ref] Coxsackie virus B4 pericarditis in mice, Tsui [/bib_ref]. Interleukin-33 (IL-33) induced eosinophilic pericarditis has also been implicated in Coxsackie B infection [bib_ref] IL-33 independently induces eosinophilic pericarditis and cardiac dilation: ST2 improves cardiac function, Abston [/bib_ref]. IFN-γ and TGF-β knockout (KO) mice bear gross histological and haemodynamic characteristics of pericarditis. Pericarditis in the IFN-γ mice presented as a thick and stiff pericardium which formed adhesions to surrounding structures. Mesothelial hyperplasia in the pericardium was accompanied by a morphological change to a cuboidal shape. In addition to the predominant mononuclear cell infiltration, the pericardial inflammatory infiltrate of the IFN-γ KO mice had marked eosinophilia. Similarly, cardiac myocytes bordering areas of inflammation in the TGF-β KO mice presented with eosinophilic inclusions and contained large nucleoli [bib_ref] Novel model of constrictive pericarditis associated with autoimmune heart disease in interferon-gammaknockout..., Afanasyeva [/bib_ref] [bib_ref] Transforming growth factor-beta 1 null mice. An animal model for inflammatory disorders, Kulkarni [/bib_ref].
## The fibrotic process and
## Constrictive pericarditis
## Constrictive pericarditis
Constrictive pericarditis is a clinical syndrome, characterised by a thickened and non-compliant pericardium, which restricts cardiac filling [bib_ref] Cardiac tamponade, constrictive pericarditis, and restrictive cardiomyopathy, Goldstein [/bib_ref]. The most apparent pathological features of constrictive pericarditis are inflammation and fibrotic thickening of the thin and elastic parietal and visceral pericardial linings. The pericardium commonly bears areas of inflammation of the serosa, scarring, and fibro-calcification [bib_ref] Cardiac tamponade, constrictive pericarditis, and restrictive cardiomyopathy, Goldstein [/bib_ref]. Constrictive pericarditis may result from severe acute inflammation or recurrent less severe inflammatory events over a highly variable time course from the period of injury [bib_ref] Constrictive pericarditis--a curable diastolic heart failure, Syed [/bib_ref]. While the predictors of progression to constriction following acute pericarditis are poorly understood the incidence of constrictive pericarditis is significantly dependent on the aetiology of the pericarditis. Whilst idiopathic and viral pericarditis have a low incidence of constrictive complications (0.8/1000 person-years), tuberculous (31.7/1000 person-years) and purulent pericarditis (52.74/1000 person-years) are associated with the highest rates of progression to SDKP) and galectin-3 could be detected in tuberculous pericardial fluid. The reduction in Ac-SDKP levels in TB pericardial effusion has been suggested to contribute to the development of fibrosis associated with TB pericarditis [bib_ref] Scientific letter: Ac-SDKP (N-acetyl-seryl-aspartyl-lysyl-proline) and Galectin-3 levels in tuberculous pericardial effusion: implications..., Ntsekhe [/bib_ref]. Elevated pericardial adenosine deaminase (ADA) activity and lysozyme levels have also been associated with TB pericarditis, and are of significant value in the diagnosis of TB pericarditis [bib_ref] Diagnosing tuberculous pericarditis, Reuter [/bib_ref]. High ADA levels are also prognostic for the development of constrictive pericarditis [bib_ref] The use of adenosine deaminase and interferon-gamma as diagnostic tools for tuberculous..., Burgess [/bib_ref].
## Malignant pericarditis
Both primary (mesotheliomas, sarcomas, fibromas) and secondary (carcinoma, lymphoma, and carcinoid) neoplasms can be accompanied by pericardial inflammation. However, neoplastic pericarditis arises mostly from secondary disorders as a result of tumour spread and metastasis through lymphatic and haematogenous spread. Effusions are common in neoplastic pericarditis and can be bloody. Malignant cells can also be present in the pericardium but almost 50% of symptomatic pericarditis cases have negative cytological results for malignant cells [bib_ref] Malignancy-related pericardial effusion. 127 cases from the Roswell Park Cancer Institute, Wilkes [/bib_ref]. However, malignancies are commonly widespread when pericardial symptoms become apparent and malignant invasion of the heart and the deposition of fibrous tissue often lead to constriction [bib_ref] Secondary malignant tumors of the pericardium, Thurber [/bib_ref]. Sub-acute inflammation with lymphocytic accumulation and mesothelial hyperplasia has been described in primary pericardial mesothelioma [bib_ref] Recurrent pericarditis related to primary pericardial malignant mesothelioma, Smets [/bib_ref]. Ristić et al [bib_ref] Pericardial cytokines in neoplastic, autoreactive, and viral pericarditis, Ristić [/bib_ref] measured elevated serum and pericardial levels of IL-6 and TGF-β in malignant pericarditis as compared to bypass surgery controls. TNF-α and IFN-γ levels were however not affected. A study by Pankuweit et al [bib_ref] Cytokine activation in pericardial fluids in different forms of pericarditis, Pankuweit [/bib_ref] , found that IFN-γ levels in effusions were found to be slightly lower than in the serum. In accordance with Ristić et al [bib_ref] Cytokine activation in pericardial fluids in different forms of pericarditis, Pankuweit [/bib_ref] IL-6 and IL-8 levels were markedly increased in pericardial fluid as compared to the serum, suggesting a local initiation of the inflammatory response. Cardiac embryonic antigen is useful in the diagnosis of malignant pericarditis and levels above 5 ng/mL are found in the majority of cases [bib_ref] Carcinoembryonic antigens in the pericardial fluid of patients with malignant pericarditis, Tatsuta [/bib_ref].
## Animal models of pericarditis
Several animal models of pericarditis have been described whereby the onset of inflammation was triggered by diverse mechanisms. A severe inflammatory reaction has been described in the pericardium of sheep injected with a bacterial toxin and Freund's adjuvant. A cellular mesothelial response was observed with changes to the morphology and disturbance to the architecture, followed by detachment from neighbouring cells and desquamation. An accompanying increase in vascular permeability resulted in the accumulation of large numbers of inflammatory cells and the exudation of fibrin. An increased collagen turnover was apparent after 6 d and the appearance of adhesions occurred as early as 2 wk post injection [bib_ref] Animal model of acute pericarditis and its progression to pericardial fibrosis and..., Leak [/bib_ref].
Ramasamy V et al . Constricitve pericaridis molecular mechanisms effusive constrictive pericarditis (associated with a high incidence of pericarditis), demonstrate a mixed picture of both pro-inflammatory IFN-γ, and anti inflammatory cytokines IL-8, and IL-10 [bib_ref] Interferon-gamma protects against chronic viral myocarditis by reducing mast cell degranulation, fibrosis,..., Fairweather [/bib_ref] , but their exact roles are as yet unclear.
Patterns of inflammation and fibrosis in the pericardium suggest that both myocardial and pericardial cells play a role in the pathogenesis of pericarditis and constriction. A change in mesothelial cell morphology has been consistently described in various forms of pericarditis. Further, a loss of the mesothelial cell architecture, as well as mesothelial desquamation often accompanies constrictive pericarditis [fig_ref] Figure 1: Molecular mediators involved in the inflammatory and fibrotic processes arising in constrictive... [/fig_ref]. The transition from a "flat" to a "cuboidal" shape has been associated with an "activation" of mesothelial cells and a distinct enzymatic profile of the cells with functions being geared towards oxidative stress and inflammatory responses [bib_ref] Physiology of pericardial fluid production and drainage, Vogiatzidis [/bib_ref] [bib_ref] The mesothelium: a cytochemical study of "activated" mesothelial cells, Whitaker [/bib_ref]. Activated mesothelial cells secrete chemokines and adhesion molecules to aid in the recruitment and migration of leukocytes across the mesothelium. They are also known to mediate the inflammatory process and produce ECM components [bib_ref] Mesothelial cells in tissue repair and fibrosis, Mutsaers [/bib_ref]. Further, mesothelial cells can undergo phenotypic changes similar to epithelial-to-mesenchymal transition to adopt fibroblast-like morphology and function in the healing serosa [bib_ref] Mesothelial cells: their structure, function and role in serosal repair, Mutsaers [/bib_ref] [bib_ref] Peritoneal dialysis and epithelial-to-mesenchymal transition of mesothelial cells, Yáñez-Mó [/bib_ref]. The active regulation of both pro-and anti-inflammatory mediators by mesothelial cells suggests a key role for the cells in maintaining pericardial homeostasis and also in the pathogenesis of pericardial fibrosis. Pericardial interstitial cells (PICs) have also been implicated in the production of ECM and calcification in the pericardium [bib_ref] Characteristics of pericardial interstitial cells and their implications in pericardial fibrocalcification, Liu [/bib_ref].
PICs have a comparable immune-phenotype to pericardial constriction [bib_ref] Evaluation and Treatment of Pericarditis: A Systematic Review, Imazio [/bib_ref].
## Molecular mechanisms of fibrosis
Although the molecular processes leading to fibrogenesis are likely to be unique for different pathologies, some key mechanisms and pathways are common to most fibrotic events [bib_ref] TGF-beta signaling and the fibrotic response, Leask [/bib_ref]. The fibrotic cascade of events is triggered upon insult to epithelial or endothelial cells which results in the activation of the coagulation cascade. The initial inflammatory response is characterised by the release of various pro-inflammatory cytokines, including tumour necrosis factor-α, TNF-α [bib_ref] Common and unique mechanisms regulate fibrosis in various fibroproliferative diseases, Wynn [/bib_ref] [bib_ref] Cellular and molecular mechanisms of fibrosis, Wynn [/bib_ref]. TNF-α is a pleiotropic cytokine with a central role in the activation and recruitment of immune cells and the regulation of pro-inflammatory cytokine production [bib_ref] Tumor necrosis factor-α signaling in macrophages, Parameswaran [/bib_ref]. Activated leukocytes then proceed to release pro-fibrotic cytokines such as IL-13 and TGF-β which drive EMT and ECM component production. TGF-β, is a key mediator of the fibrotic response and it acts via canonical (Smaddependent) and non-canonical (non-Smad-based) signalling pathways to coordinate an ECM accumulation through increased synthesis as well as a decreased degradation of ECM components [bib_ref] TGF-beta and fibrosis, Branton [/bib_ref] [bib_ref] Smad regulation in TGF-beta signal transduction, Moustakas [/bib_ref] [bib_ref] Non-Smad pathways in TGF-beta signaling, Zhang [/bib_ref].
## Molecular mechanisms of pericardial fibrosis
Molecular mechanisms of pericardial constriction remain to be fully elucidated but are likely to follow a classical pattern of pericardial inflammation mediated by various cytokines , including TNF-α, followed by abnormal healing with an exaggerated TGF-β mediated profibrotic response leading to pericardial fibrosis. Both experimental mice models of acute pericarditis and pericardial fluid from patients with tuberculous Summary of Inflammatory and fibrotic cytokines and growth factors (detected in pericardial fluid) likely to modulate the pathophysiological processes leading to chronic fibrosis in the pericardium TGF-β: Growth factor β; CTGF: Connective tissue growth factor; TNF-α: Tumor necrosis factor-α; IFN-γ: Interferon-γ; VEGF: Vascular endothelial growth factor; bFGF: Basic fibroblast growth factor; ECM: Extracellular matrix. Ramasamy V et al . Constricitve pericaridis molecular mechanisms infection, and is believed to arise from an overflow of ACE production by the macrophages and phagocytes in the granulomatous lesions into the circulation. Thus increased ACE levels, resulting in increased enzymatic cleavage of Ac-SDKP and a subsequent dampening in its anti-fibrotic potential, could potentially contribute to the pathogenesis of constriction. Finally, an increase in CTGF was associated with ECM deposition and pericardial remodelling [bib_ref] Pathogenetic mechanisms in radiation fibrosis, Yarnold [/bib_ref]. This is not surprising as CTGF expression is known to be induced by TGF-β in cardiac fibroblasts and cardiac myocytes, whereby it contributes to the expression of fibronectin, collagen type Ⅰ and plasminogen activator inhibitor-1 [bib_ref] CTGF expression is induced by TGF-beta in cardiac fibroblasts and cardiac myocytes:..., Chen [/bib_ref].
## Inflammatory/ fibrotic mediator major roles in inflammation and fibrosis references
Interestingly a decrease in VEGF was observed in viral pericarditis which rarely results in a constrictive pericarditis. Whilst VEGF mediated angiogenesis is known to be important for the promotion of fibrosis, it also plays a role in fibrosis resolution [bib_ref] Vascular endothelial growth factor promotes fibrosis resolution and repair in mice, Yang [/bib_ref]. Indeed, an angio-fibrotic switch of VEGF and CTGF has been described in proliferative diabetic retinopathy, whereby the VEGF to CTGF ratio closely dictates the progression to fibrosi [bib_ref] A shift in the balance of vascular endothelial growth factor and connective..., Van Geest [/bib_ref] [bib_ref] The angio-fibrotic switch of VEGF and CTGF in proliferative diabetic retinopathy, Kuiper [/bib_ref]. CTGF has also been shown to bind to VEGF and to inhibit its angiogenic functions [bib_ref] Connective tissue growth factor binds vascular endothelial growth factor (VEGF) and inhibits..., Inoki [/bib_ref]. Hence, it is possible that such CTGF-VEGF interplay is also involved in the progression to fibrosis in the pericardium. This would further explain the high VEGF levels coinciding with low levels of bFGF in viral pericardial fluid.
# Conclusion
In this review, we have highlighted that the pericardium is subjected to noxious injury from a wide spectrum of infections and non infections causes and that the pathogenesis of pericarditis in each instance may differ in significant ways. Importantly the progression from mesenchymal stem cells. PICs cultured from fibrocalcific human samples could be differentiated into myofibroblasts and osteoblasts which are central to the development of fibrosis and the production of extraosseous calcification. TGF-β and bone morphogenetic protein 2 (BMP-2) were associated with the transdifferentiation process. TGF-β increased PIC mRNA expression of collagens Ⅰ and Ⅲ whilst decreasing the matrix metalloprotease-2 and -9 mRNA levels which are important for elastin degradation, thus regulating the fibrotic process by modulating fibrosis related gene expression [bib_ref] Characteristics of pericardial interstitial cells and their implications in pericardial fibrocalcification, Liu [/bib_ref]. TGF-β is a master regulator of extracellular matrix component expression and the development of fibrosis. Increased pericardial fluid and serum levels of TGF-β have been described in various forms of pericarditis and have been associated with increased collagen synthesis. It is thus safe to say that TGF-β might play a key role in the development of fibrosis in the pericardium. Interestingly, Ristic et al [bib_ref] Pericardial cytokines in neoplastic, autoreactive, and viral pericarditis, Ristić [/bib_ref] did not detect any increase in TGF-β levels in viral pericardial fluid and this could account for the low proportion of constrictive pericarditis arising in this group. Ac-SDKP, which is known to decrease TGF-β signalling by decreasing TGF-β transcription and the phosphorylation of Smad2 and Smad3 and their translocation to the nucleus [bib_ref] Prevention of myocardial fibrosis by N-acetyl-seryl-aspartyl-lysyl-proline in diabetic rats, Castoldi [/bib_ref] [bib_ref] Increased myocardial collagen content in transgenic rats overexpressing cardiac angiotensin-converting enzyme is..., Pokharel [/bib_ref] [bib_ref] Prevention of aortic fibrosis by N-acetyl-serylaspartyl-lysyl-proline in angiotensin II-induced hypertension, Lin [/bib_ref] , may play a role in the pathway to pericardial fibrosis. Patients with TB pericarditis, have been found to have diminished Ac-SDKP levels compared to particpants without pericarditis undergoing cardiac sugery [bib_ref] Scientific letter: Ac-SDKP (N-acetyl-seryl-aspartyl-lysyl-proline) and Galectin-3 levels in tuberculous pericardial effusion: implications..., Ntsekhe [/bib_ref]. Lowered Ac-SDKP levels could arise from an increase in angiotensin converting enzyme (ACE) activity, which is known to degrade Ac-SDKP [bib_ref] Angiotensin I-converting enzyme and metabolism of the haematological peptide N-acetyl-seryl-aspartyl-lysyl-proline, Azizi [/bib_ref] [bib_ref] Quantification of N-acetyl-seryl-aspartyl-lysyl-proline in hemodialysis patients administered angiotensin-converting enzyme inhibitors by stable..., Inoue [/bib_ref]. An increase in ACE serum levels has been reported in granulomatous conditions including Mtb pericarditis, to the development of pericardial fibrosis varies significantly by etiology. A classical pattern of inflammation in the pericardium mediated by various cytokines is likely to occur as a result of most types of insult. However, the unfolding of events leading to the development of fibrosis post-inflammation is harder to accurately predict. Nevertheless, this review has allowed us to postulate various cellular events and signalling cascades which are likely to contribute, albeit to different extents in varying types of pericarditis, to the pathological fibrotic phenotype. Whilst the role of common players such as TGF-β and TNF-α in the inflammatory process can be quite easily predicted, their complex range of functions makes them unattractive targets in the management and treatment of constrictive pericarditis. However, the identification of other pro-and anti-fibrotic mediators such as Galectin-3, Ac-SDKP and bFGF, with a narrower range of functions could represent new avenues for the treatment of pericarditis. Research aimed at developing a better understanding of molecular mechanisms involved in the progression of pericarditis to fibrosis may be able to (1) identify high risk patients for progression to constrictive pericarditis through novel markers of fibrosis; and (2) identify novel targets for therapy to interrupt the progression to fibrosis and prevent the development of constrictive pericarditis.
[fig] Figure 1: Molecular mediators involved in the inflammatory and fibrotic processes arising in constrictive pericarditis. Ac-SDKP: N-acetyl-seryl-aspartyl-lysylproline; TGF-β: Growth factor β; TNF-α: Tumor necrosis factor-α; CTGF: Connective tissue growth factor; IFN-γ: Interferon-γ; VEGF: Vascular endothelial growth factor. [/fig]
|
Monitoring Neurochemistry in Traumatic Brain Injury Patients Using Microdialysis Integrated with Biosensors: A Review
# Introduction
Traumatic brain injury (TBI) is a health burden affecting over 50 million people each year globally [bib_ref] Traumatic brain injury: Integrated approaches to improve prevention, clinical care, and research, Maas [/bib_ref] [bib_ref] Estimating the global incidence of traumatic brain injury, Dewan [/bib_ref]. Its long-term adverse effects to the brain may create a significant burden implicating social integration, independence, and employability. Populationbased studies have reported that 238,000 people each year sustain a TBI in the UK, and it is the overall leading cause of death in young adults in developed countries [bib_ref] Traumatic brain injury, Barlow [/bib_ref]. In 2013, TBIs in the USA led to about 2.5 million emergency department visits, 282,000 hospitalisations, and 56,000 deaths [bib_ref] Traumatic Brain Injury-Related Emergency Department Visits, Hospitalizations, and Deaths-United States, Taylor [/bib_ref]. Mild or severe TBI patients are at risk of secondary brain injury that arises from complex physiological responses to the initial injury. These complex physiological processes are thought to include a series of sequential events including altered cerebral metabolism. Monitoring cerebral metabolism along with other protocols to identify early changes and prevent secondary brain injuries during neurocritical care management can improve outcomes of severe TBI patients [bib_ref] What's new in the management of traumatic brain injury on neuro ICU?, Wijayatilake [/bib_ref]. Currently, cerebral metabolism is monitored using cerebral microdialysis, an invasive technique in which interstitial brain fluid can be sampled. The sampled fluid (also termed microdialysate) is usually analysed hourly offline using a bedside analyser, for which its operating principle is based on enzymatic reactions [bib_ref] 13 C-labelled microdialysis studies of cerebral metabolism in TBI patients, Carpenter [/bib_ref] [bib_ref] Microdialysis in the human brain and its potential role in the development..., Helmy [/bib_ref]. However, the data analysed are retrospective, hindering real-time interventions. Thus, effective and timely measurements of the microdialysates are required to predict the onset of secondary brain damage. To fulfil this, emerging research has integrated microdialysis with biosensors for continuous online detection of relevant cerebral metabolites found in microdialysates. The intent of these technologies is to provide accurate and actionable information that may help inform clinical decision making and aid in the prognostication of patients' outcome.
In this review, the state of emerging biosensors coupled with microdialysis technology in monitoring cerebral metabolism of TBI is presented and discussed. We start by explaining the broad overview of TBI and how it alters cerebral metabolism, followed by a description of the cerebral microdialysis principle and its clinical usefulness. Recent developments in novel biosensors to monitor brain metabolism are presented, and their performances in quantifying relevant brain metabolites are also discussed. We conclude by discussing the need for the further development of novel biosensors for direct in vivo real-time monitoring of brain metabolism.
## Importance of monitoring brain metabolism for tbi
## Traumatic brain injury (tbi)
TBI is an insult to the brain due to an external force causing damage to the brain's structure and function [bib_ref] Position statement: Definition of traumatic brain injury, Menon [/bib_ref]. It represents mild, moderate, and severe effects of physical assault to the brain that may cause sequential, primary, or secondary ramifications. Thus, TBI can be classed into a primary mechanical injury, which is non-amenable to medical treatment and a delayed secondary brain injury involving various changes at the cellular and molecular level that could in theory be counteracted. The sudden mechanical injury is a widespread tearing, shearing, or stretching of axons, referred to as axonal injury, as well as contusions, hemorrhages, and lacerations [bib_ref] Traumatic Brain Injuries: Pathophysiology and Potential Therapeutic Targets. Front, Ng [/bib_ref]. Delayed secondary injury evolves hours or days after the initial primary mechanical trauma due to neuronal and glial dysfunction, neuroinflammation, cerebral oedema, and metabolic changes. Consequently, this leads to various physiologic alterations including hypoperfusion, blood-brain barrier (BBB) disruption, oxidative injury, and mitochondrial dysfunction [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref] [bib_ref] Role of extracellular glutamate measured by cerebral microdialysis in severe traumatic brain..., Chamoun [/bib_ref] [bib_ref] Monitoring of Protein Biomarkers of Inflammation in Human Traumatic Brain Injury Using..., Dyhrfort [/bib_ref] [bib_ref] Systemic inflammation alters the neuroinflammatory response: A prospective clinical trial in traumatic..., Lassarén [/bib_ref] [bib_ref] Voltage-dependent Na + /K + ion channel blockade fails to ameliorate behavioral..., Di [/bib_ref] [bib_ref] Matrix Metalloproteinase Expression in Contusional Traumatic Brain Injury: A Paired Microdialysis Study, Guilfoyle [/bib_ref]. This complexity of secondary brain injury poses diagnostic and therapeutic challenges; therefore, additional invasive monitoring interventions are required during neurocritical care of severe TBI patients. These additional metrics may provide insights into brain tissue oxygenation, intracranial pressure (ICP), cerebral perfusion pressure (CPP), electrophysiology, and local brain metabolism [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref]. Abnormal local brain metabolism is linked to poor patient clinical outcomes, thus interrogating the brain to monitor its chemistry is necessary [bib_ref] Cerebral extracellular chemistry and outcome following traumatic brain injury: A microdialysis study..., Timofeev [/bib_ref].
## Cerebral metabolism
Evidence shows that cerebral metabolism is disturbed following TBI, although the exact mechanisms are incompletely understood due to the complex and heterogeneous nature of TBI [bib_ref] 13 C-labelled microdialysis studies of cerebral metabolism in TBI patients, Carpenter [/bib_ref]. The brain uses glucose as a preferred substrate for energy consumption, so the regulation of cerebral glucose metabolism is crucial. Oxidative metabolism of glucose provides most of the ATPs utilised by the brain. However, biosynthetic routes that branch from the glycolytic pathway and the tricarboxylic acid (TCA) cycle and other pathways including the pentose phosphate shunt, glucose storage as glycogen, and the malate-aspartate shuttle all have significant roles [bib_ref] Brain Glucose Metabolism: Integration of Energetics with Function, Dienel [/bib_ref]. Understanding of the altered cerebral metabolism is incomplete without knowledge of the glucose metabolism in an uninjured brain. presents the major energy pathways in the brain.
In the uninjured brain, glucose uptake by both neuronal and glial cells undergoes oxidation via glycolysis to form two pyruvate molecules. Glycolysis is independent of oxygen and, thus, can occur under either aerobic or anaerobic conditions. The energy yield of ATP per molecule of glucose depends on whether mitochondrial shuttle mechanisms are operational or not. To generate one molecule of pyruvate by glycolysis, one molecule of NAD + is converted into NADH, which must be recycled (oxidised) back to NAD + if glycolysis is to be sustained. One recycling mechanism is by the electron-transport chains of mitochondria (if operational). As NADH cannot cross the mitochondrial membrane, the requisite hydrogens and electrons are transferred indirectly by "shuttle" mechanisms-malate-aspartate shuttle system and/or the glycerol-3-phosphate shuttle [bib_ref] Neuronal and astrocytic shuttle mechanisms for cytosolicmitochondrial transfer of reducing equivalents: Current..., Mckenna [/bib_ref]. NADH can also be recycled to NAD + by the lactate dehydrogenase (LDH)-mediated conversion of pyruvate to lactate in the cytosol, as an epilogue to "anaerobic" glycolysis.
etabolites 2022, 12, x FOR PEER REVIEW . Illustration of glucose, lactate, and pyruvate within the major energy p nases uses ATP to phosphorylate glucose to glucose 6P (glucose 6-phosphate) in t step of glycolysis. Glucose 6P can either undergo alternative metabolic fates or c glycolytic pathway to generate pyruvate. Pyruvate can be utilised either by oxi via the tricarboxylic acid (TCA) cycle or remain in the cytosol where it is conve glycolytic pathway that takes place in the cytoplasm produces a net yield of 2 A glucose (2 ATP molecules are utilised early on but then paid back later with t ATP). Overall, the yield per mole of glucose metabolised fully to CO2 by combin and mitochondrial metabolism is theoretically 36-38 moles of ATP.
In the uninjured brain, glucose uptake by both neuronal and glial . Illustration of glucose, lactate, and pyruvate within the major energy pathways. Hexokinases uses ATP to phosphorylate glucose to glucose 6P (glucose 6-phosphate) in the first irreversible step of glycolysis. Glucose 6P can either undergo alternative metabolic fates or continue down the glycolytic pathway to generate pyruvate. Pyruvate can be utilised either by oxidative metabolism via the tricarboxylic acid (TCA) cycle or remain in the cytosol where it is converted to lactate. The glycolytic pathway that takes place in the cytoplasm produces a net yield of 2 ATP per molecule of glucose (2 ATP molecules are utilised early on but then paid back later with the generation of 4 ATP). Overall, the yield per mole of glucose metabolised fully to CO 2 by combination of glycolysis and mitochondrial metabolism is theoretically 36-38 moles of ATP. The overall reaction of glycolysis and net yield of ATP molecules is presented in Equation [bib_ref] Traumatic brain injury: Integrated approaches to improve prevention, clinical care, and research, Maas [/bib_ref].
[formula] Glucose + 2ADP + 2P i + 2N AD + → 2pyruvate + 2ATP + 2N ADH + 2H + + 2H 2 O(1) [/formula]
Pyruvate, the end-product of glycolysis, can enter mitochondria where it is converted to acetyl-CoA by pyruvate dehydrogenase. Acetyl-CoA is further metabolised in the tricarboxylic acid (TCA) cycle in mitochondria. The sum of all the reactions in the TCA cycle is presented in Equation (2).
[formula] Acetyl − CoA + 3N AD + + 3FAD + GDP + P i + 2H 2 O → CoA − SH + 3N ADH + FADH 2 + 3H + + GTP + 2CO 2 (2) [/formula]
Subsequently, NADH and FADH 2 are utilised by the mitochondrial electron transport chain (by Complexes I and II, respectively); electrons are transferred to complexes III and IV, where O 2 is the terminal electron acceptor on Complex IV, followed by ATP synthesis by ATP synthase (also termed Complex V) [bib_ref] The mitochondrial respiratory chain, Rich [/bib_ref]. The yield per molecule of glucose metabolised fully to CO 2 (by combined glycolysis, NADH shuttling, and mitochondrial respiration) is theoretically 36-38 ATP molecules. However, the actual yield is considered somewhat lower.
## Altered cerebral metabolism due to tbi
Imbalance in cerebral glucose metabolism following TBI is well-documented [bib_ref] 13 C-labelled microdialysis studies of cerebral metabolism in TBI patients, Carpenter [/bib_ref] [bib_ref] Glycolysis and the Pentose Phosphate Pathway after Human Traumatic Brain Injury: Microdialysis..., Jalloh [/bib_ref] [bib_ref] A Comparison of Oxidative Lactate Metabolism in Traumatically Injured Brain and Control..., Jalloh [/bib_ref] and attributable, at least partly, to the altered ATP production in the brain's major energy pathways. Due to a high energy demand following TBI, abnormally low levels (<0.8 mM) of extracellular glucose occur [bib_ref] Cerebral extracellular chemistry and outcome following traumatic brain injury: A microdialysis study..., Timofeev [/bib_ref] , possibly because of upregulated glucose uptake by neurones and glia. Conversely, neurones and glia may sometimes be too damaged to take up glucose from extracellular fluid, leading to hypometabolism characterised by abnormally high extracellular glucose. Thus, there is an optimum extracellular glucose range, although there is insufficient evidence to define this exactly [bib_ref] Consensus statement from the 2014 International Microdialysis Forum, Hutchinson [/bib_ref]. Extracellular lactate can also be utilised as an alternative fuel [bib_ref] A Comparison of Oxidative Lactate Metabolism in Traumatically Injured Brain and Control..., Jalloh [/bib_ref]. [bib_ref] Monitoring of Protein Biomarkers of Inflammation in Human Traumatic Brain Injury Using..., Dyhrfort [/bib_ref] C-labelled microdialysis studies have demonstrated that the traumatically injured brain uses lactate via the TCA cycle [bib_ref] A Comparison of Oxidative Lactate Metabolism in Traumatically Injured Brain and Control..., Jalloh [/bib_ref] [bib_ref] The human brain utilizes lactate via the tricarboxylic acid cycle: A 13C-labelled..., Gallagher [/bib_ref]. Another 13 C-labelled microdialysis study found lactate production from 1,2-13 C 2 glucose via glycolysis and to a lesser extent via PPP [bib_ref] Glycolysis and the Pentose Phosphate Pathway after Human Traumatic Brain Injury: Microdialysis..., Jalloh [/bib_ref]. Lactate was also identified as a spin-out product (cataplerosis) from the TCA cycle in 13 C-labelled microdialysis studies using 2,3-13 C 2 succinate as a substrate [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref] [bib_ref] Glycolysis and the Pentose Phosphate Pathway after Human Traumatic Brain Injury: Microdialysis..., Jalloh [/bib_ref].
A persistent high lactate/pyruvate ratio (LPR) (LPR > 25 or >40) indicates metabolic dysfunction or crisis. In a microdialysis study of 233 TBI patients, acute-phase LPR > 25 was associated with poor clinical outcomes 6 months later [bib_ref] Cerebral extracellular chemistry and outcome following traumatic brain injury: A microdialysis study..., Timofeev [/bib_ref]. High LPR, despite seemingly adequate oxygen and glucose delivery to brain tissues, is regarded as indicating mitochondrial dysfunction [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref] [bib_ref] Characterising the dynamics of cerebral metabolic dysfunction following traumatic brain injury: A..., Guilfoyle [/bib_ref]. The concentrations of lactate and pyruvate and their ratio (LPR) provide useful information about the cellular redox state in the region of interest. The extracellular LPR is thought to reflect the LPR in the cytoplasm-itself in equilibrium with cytoplasmic NADH/NAD + ratio [bib_ref] The redox state of free nicotinamide-adenine dinucleotide in the cytoplasm and mitochondria..., Williamson [/bib_ref]. Glucose, lactate, pyruvate, and LPR were cited as the most clinically relevant biomarkers in a consensus statement from the 2014 International Microdialysis Forum [bib_ref] Consensus statement from the 2014 International Microdialysis Forum, Hutchinson [/bib_ref]. Timely assessment of these is, therefore, essential in the early detection of secondary brain injury allowing prompt interventions. summarises neuroprotective interventions for altered neurochemistry. . Summary of neuroprotective interventions for altered cerebral metabolism [bib_ref] Intensive insulin therapy reduces microdialysis glucose values without altering glucose utilization or..., Vespa [/bib_ref] [bib_ref] Tight glycemic control increases metabolic distress in traumatic brain injury: A randomized..., Vespa [/bib_ref] [bib_ref] Impact of tight glycemic control on cerebral glucose metabolism after severe brain..., Oddo [/bib_ref] [bib_ref] Systemic glucose and brain energy metabolism after subarachnoid hemorrhage, Helbok [/bib_ref] [bib_ref] Monitoring of glucose in brain, adipose tissue, and peripheral blood in patients..., Rostami [/bib_ref] [bib_ref] Normobaric hyperoxia-induced improvement in cerebral metabolism and reduction in intracranial pressure in..., Tolias [/bib_ref] [bib_ref] Effect of hyperoxia on regional oxygenation and metabolism after severe traumatic brain..., Nortje [/bib_ref] [bib_ref] A prospective, randomized Phase II clinical trial to evaluate the effect of..., Rockswold [/bib_ref] [bib_ref] Effect of hyperventilation on extracellular concentrations of glutamate, lactate, pyruvate, and local..., Marion [/bib_ref] [bib_ref] Correlation between Cerebral Blood Flow, Substrate Delivery, and Metabolism in Head Injury:..., Hutchinson [/bib_ref] [bib_ref] Effects of mannitol bolus administration on intracranial pressure, cerebral extracellular metabolites, and..., Sakowitz [/bib_ref] [bib_ref] Flying in silence: Echolocating bats cease vocalizing to avoid sonar jamming, Chiu [/bib_ref] [bib_ref] Relationship between brain temperature, brain chemistry and oxygen delivery after severe human..., Soukup [/bib_ref] [bib_ref] Effects of hypothermia on excitatory amino acids and metabolism in stroke patients:..., Berger [/bib_ref] [bib_ref] Cerebral oxygenation, vascular reactivity, and neurochemistry following decompressive craniectomy for severe traumatic..., Ho [/bib_ref] [bib_ref] Relevance of intracranial hypertension for cerebral metabolism in aneurysmal subarachnoid hemorrhage. Clinical..., Nagel [/bib_ref].
## Intervention effect references
Glucose/insulin ↑↓ glucose, ↑↓ LPR [bib_ref] Intensive insulin therapy reduces microdialysis glucose values without altering glucose utilization or..., Vespa [/bib_ref] [bib_ref] Tight glycemic control increases metabolic distress in traumatic brain injury: A randomized..., Vespa [/bib_ref] [bib_ref] Impact of tight glycemic control on cerebral glucose metabolism after severe brain..., Oddo [/bib_ref] [bib_ref] Systemic glucose and brain energy metabolism after subarachnoid hemorrhage, Helbok [/bib_ref] [bib_ref] Monitoring of glucose in brain, adipose tissue, and peripheral blood in patients..., Rostami [/bib_ref] Hyperoxia ↑ PBtO 2 , variable ↓ LPR [bib_ref] Normobaric hyperoxia-induced improvement in cerebral metabolism and reduction in intracranial pressure in..., Tolias [/bib_ref] [bib_ref] Effect of hyperoxia on regional oxygenation and metabolism after severe traumatic brain..., Nortje [/bib_ref] [bib_ref] A prospective, randomized Phase II clinical trial to evaluate the effect of..., Rockswold [/bib_ref] Hyperventilation ↓ glucose [bib_ref] Effect of hyperventilation on extracellular concentrations of glutamate, lactate, pyruvate, and local..., Marion [/bib_ref] [bib_ref] Correlation between Cerebral Blood Flow, Substrate Delivery, and Metabolism in Head Injury:..., Hutchinson [/bib_ref] Mannitol ↓ LPR [bib_ref] Systemic glucose and brain energy metabolism after subarachnoid hemorrhage, Helbok [/bib_ref] [bib_ref] Effects of mannitol bolus administration on intracranial pressure, cerebral extracellular metabolites, and..., Sakowitz [/bib_ref] Decompressive craniotomy ↓ LPR [bib_ref] Cerebral oxygenation, vascular reactivity, and neurochemistry following decompressive craniectomy for severe traumatic..., Ho [/bib_ref] [bib_ref] Relevance of intracranial hypertension for cerebral metabolism in aneurysmal subarachnoid hemorrhage. Clinical..., Nagel [/bib_ref] Therapeutic (induced) hypothermia ↓ glucose, ↓ lactate [bib_ref] Relationship between brain temperature, brain chemistry and oxygen delivery after severe human..., Soukup [/bib_ref] [bib_ref] Effects of hypothermia on excitatory amino acids and metabolism in stroke patients:..., Berger [/bib_ref]
## Review of sensor technologies for brain metabolism
In selecting scholarly articles to discuss in this review, emphasis was placed on work published in the areas of biosensors, TBI, cerebral metabolism, and microdialysis in the last 10 years. We first discuss the current routinely used method and then present others. [fig_ref] Table 2: A summary comparison of recent key examples of biosensors integrated with microdialysis... [/fig_ref] summarises some of the most relevant articles discussed in this review. Those biosensors are in the research and development phases and are not approved for routine clinical use.
## Cerebral microdialysis
The use of cerebral microdialysis during neurocritical care management of TBI patients along with other metrics such as ICP and PbtO 2 aims at minimising or preventing the burden of secondary brain damage [bib_ref] Consensus statement from the 2014 International Microdialysis Forum, Hutchinson [/bib_ref] [bib_ref] Characterising the dynamics of cerebral metabolic dysfunction following traumatic brain injury: A..., Guilfoyle [/bib_ref]. Cerebral microdialysis is a neuromonitoring tool that reflects brain-tissue health and metabolism [bib_ref] Microdialysis in traumatic brain injury-methodology and pathophysiology, Hutchinson [/bib_ref]. A microdialysis probe perfused with physiological salts solution is placed in the brain's parenchymal region, allowing sampling of molecules from the brain's extracellular fluid. The cerebral microdialysis technique including its components is represented in [fig_ref] Figure 2: Cerebral microdialysis [/fig_ref].
bolites 2022, 12, x FOR PEER REVIEW 6 of perfused with physiological salts solution is placed in the brain's parenchymal regio allowing sampling of molecules from the brain's extracellular fluid. The cerebral micro alysis technique including its components is represented in [fig_ref] Figure 2: Cerebral microdialysis [/fig_ref]. , and brain tissue oxygen (O2) probes through the skull in the brain (image copyright K.L.H. Carpenter, reproduced with her permission). A syringe for su plying perfusion fluid and a microdialysate collection vial are also shown. A pump (not show drives the syringe that delivers perfusion fluid into the microdialysis catheter. The microdialy catheter tip is typically placed in the white matter of the right frontal lobe, as shown in the magne resonance image (MRI) (B) (reproduced from [bib_ref] 13 C-labelled microdialysis studies of cerebral metabolism in TBI patients, Carpenter [/bib_ref]. (C) The architecture of the microdialysis cathe tip (adapted from [bib_ref] Cerebral microdialysis in clinical studies of drugs: Pharmacokinetic applications, Shannon [/bib_ref]. The microdialysates are assayed offline, typically hourly, using the CUSflex bedside analyser (D) (image reproduced from [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref] for glucose, lactate, and pyruvate a often also for glutamate and glycerol. Images (A,B,D) originally appeared in [bib_ref] Cerebral microdialysis in clinical studies of drugs: Pharmacokinetic applications, Shannon [/bib_ref] , [bib_ref] 13 C-labelled microdialysis studies of cerebral metabolism in TBI patients, Carpenter [/bib_ref] , and [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref] spectively, published Open Access CC BY, copyright The Authors.
In a clinical setting, cerebral microdialysis is used along with ICP and PBtO2 senso and is a part of multi-modality monitoring enabled by a triple lumen cranial access dev , and brain tissue oxygen (O 2 ) probes through the skull into the brain (image copyright K.L.H. Carpenter, reproduced with her permission). A syringe for supplying perfusion fluid and a microdialysate collection vial are also shown. A pump (not shown) drives the syringe that delivers perfusion fluid into the microdialysis catheter. The microdialysis catheter tip is typically placed in the white matter of the right frontal lobe, as shown in the magnetic resonance image (MRI) (B) (reproduced from [bib_ref] 13 C-labelled microdialysis studies of cerebral metabolism in TBI patients, Carpenter [/bib_ref]. (C) The architecture of the microdialysis catheter tip (adapted from [bib_ref] Cerebral microdialysis in clinical studies of drugs: Pharmacokinetic applications, Shannon [/bib_ref]. The microdialysates are assayed offline, typically hourly, using the ISCUSflex bedside analyser (D) (image reproduced from [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref] for glucose, lactate, and pyruvate and often also for glutamate and glycerol. Images (A,B,D) originally appeared in [bib_ref] Cerebral microdialysis in clinical studies of drugs: Pharmacokinetic applications, Shannon [/bib_ref] , [bib_ref] 13 C-labelled microdialysis studies of cerebral metabolism in TBI patients, Carpenter [/bib_ref] , and [bib_ref] Recent advances in traumatic brain injury, Khellaf [/bib_ref] respectively, published Open Access CC BY, copyright The Authors.
In a clinical setting, cerebral microdialysis is used along with ICP and PBtO 2 sensors and is a part of multi-modality monitoring enabled by a triple lumen cranial access device [bib_ref] Clinical cerebral microdialysis: A methodological study, Hutchinson [/bib_ref] [fig_ref] Figure 2: Cerebral microdialysis [/fig_ref] , which holds the three probes in the brain parenchyma [fig_ref] Figure 2: Cerebral microdialysis [/fig_ref]. Typically, CNS perfusion fluid (M Dialysis AB, Stockholm, Sweden), containing 147 mM NaCl, 2.7 mM KCl, 1.2 mM CaCl 2 , and 0.85 mM MgCl 2 in ultrapure water, is pumped (using a syringe pump) at a low flowrate (typically 0.3 µL/min) through the inlet tubing and down the outer tube of the catheter to the distal end where the last 10 mm (or in some cases 20 mm or 30 mm) of the outer tube is composed of a semi-permeable membrane, where the bi-directional exchange of molecules occurs by diffusion between the fluid in the catheter and the brain's extracellular interstitial space [bib_ref] Microdialysis in the human brain and its potential role in the development..., Helmy [/bib_ref].
The catheter tip has a closed end, allowing the return of the perfusion fluid-now termed microdialysate-up the inner tube of the catheter, and it is collected in a micro-vial. A schematic of the microdialysis catheter tip is shown in [fig_ref] Figure 2: Cerebral microdialysis [/fig_ref]. The semi-permeable membrane typically used clinically has either a 20 kDa or 100 kDa nominal molecular weight cut-off. The microdialysate contains the molecules of interest, including markers of brain metabolism such as glucose, lactate, and pyruvate. Mean relative recoveries (%) (also known as extraction efficiency) using 20 kDa or 100 kDa catheters at a flow rate of 0.3 µL/min have been determined in vitro as 94-105% for glucose, 87-95% for lactate, and 92-105% pyruvate [bib_ref] Microdialysis in traumatic brain injury-methodology and pathophysiology, Hutchinson [/bib_ref].
For clinical monitoring, the collected microdialysate in the microvial is typically analysed hourly, using the standard enzymatic colorimetric bedside analyser (ISCUSflex clinical microdialysis analyser, M Dialysis AB, Stockholm, Sweden), which is shown in [fig_ref] Figure 2: Cerebral microdialysis [/fig_ref]. Timely assessment of these metabolites is essential in the early detection of secondary brain injury, allowing prompt interventions. Therefore, the development of novel biosensors for neurochemical monitoring is centred around the rapid and continuous detection of the relevant metabolites in TBI patients.
## Current standard for tbi monitoring in the icu
The concentrations of glucose, lactate, and pyruvate in patients' brain extracellular fluid samples retrieved through microdialysis have been determined through a widely used ISCUSflex Microdialysis Analyser (M Dialysis AB, Stockholm, Sweden). The commercially available ISCUSflex analyser and, from the same manufacturer, earlier models, ISCUS and CMA600, are examples of enzymatic colorimetric analysers. Glucose, lactate, and pyruvate are each enzymatically oxidised by reagents that contain glucose oxidase, lactate oxidase, and pyruvate oxidase, respectively, producing hydrogen peroxide as a by-product. The reagents also contain peroxidase and dye-precursors, and when they react with the hydrogen peroxide, this leads to the formation of redviolet-coloured products: quinoneimines for glucose and lactate, and a quinonediimine for pyruvate. The rate of formation of these coloured products is proportional to concentrations of each analyte. Glucose, lactate, and pyruvate are assayed sequentially on discrete small aliquots of microdialysates (0.5, 0.2, and 0.5 mL, respectively). IS-CUSflex, ISCUS, and CMA600 are advantageous due to their ability to handle small microdialysate volumes and have low limits of detection of 0.1 mM for both glucose and lactate and 10 mM for pyruvate. The ability of the ISCUSflex, ISCUS, and CMA600 to produce absolute values of the three clinically relevant metabolites along with LPR has proven useful in microdialysis-based studies where findings have been correlated with clinical outcomes [bib_ref] Cerebral extracellular chemistry and outcome following traumatic brain injury: A microdialysis study..., Timofeev [/bib_ref] [bib_ref] Characterising the dynamics of cerebral metabolic dysfunction following traumatic brain injury: A..., Guilfoyle [/bib_ref]. Though the ISCUSflex, ISCUS and CMA600 satisfy the clinical requirements with respect to sensitivity and selectivity, and they are hampered by the need for fresh reagents every few days and these are expensive. They also require hourly analyses by clinical staff, and, therefore, are very labour-intensive. On a busy shift, clinical staff are sometimes unable to analyse microdialysates in a timely manner. The limitation of hourly-or even less frequent-readings lead, in effect, to retrospective metabolic assessments that are less useful to making clinical judgements. These drawbacks highlight the need for devising biosensing techniques for online continuous monitoring of brain metabolism in TBI patients during neurocritical care. It should be noted that ISCUS and CMA600 are no longer marketed; ISCUSflex [fig_ref] Figure 2: Cerebral microdialysis [/fig_ref] is the only clinical enzymatic colorimetric microdialysis analyser currently available commercially.
# Introduction to biosensors
By design, any biosensor comprises (1) an element that detects the substrate of interest and produces a response signal, (2) a transducer, which transforms the generated response into a detectable response, and (3) a detector capable of enhancing and processing signals before they are displayed on an electronic digital platform [bib_ref] Introduction to biosensors, Bhalla [/bib_ref]. It is noteworthy that the field of biosensors is very broad, and their categorisation in the literature is not straightforward. However, based on a transducer, biosensors can be classified as electrochemical, optical, and others [bib_ref] Introduction to biosensors, Bhalla [/bib_ref] [bib_ref] Recent advances of electrochemical and optical enzyme-free glucose sensors operating at physiological..., Adeel [/bib_ref] [bib_ref] Simultaneous optical and electrochemical label-free biosensing with ITO-coated lossy-mode resonance sensor, Śmietana [/bib_ref] [bib_ref] The role of biosensors in the detection of emerging infectious diseases, Pejcic [/bib_ref].
Ever since Clark and Lyons introduced a biosensor to monitor blood gas levels in 1962 [bib_ref] The future of biosensors, Bergveld [/bib_ref] , recent advances have facilitated alternative detection modalities. Since then, various types of biosensors for specific detection of analytes [bib_ref] Prostate cancer detection: A systematic review of urinary biosensors, Chan [/bib_ref] [bib_ref] Biosensors and nanotechnology for cancer diagnosis (lung and bronchus, breast, prostate, and..., Sharifianjazi [/bib_ref] have been developed; hence, this is now a multidisciplinary area bridging fundamental science (chemistry, physics, and biology) and clinical practice. An early example of a biosensor integrated with microdialysis for diagnostics was reported by Roda and colleagues in 1991 to monitor extracellular lactate in a rat brain [bib_ref] Microdialysis and luminescent probe: Analytical and clinical aspects, Roda [/bib_ref]. The enzyme-based electrochemical biosensor showed a high performance in quantifying lactate attributed to a limit of detection (LoD) of 0.1 mM, a wide linearity range between 0.1 and 100 mM, and an imprecision of less than 5%. Advantageously, this biosensor quantified other small molecules including glycerol. Since then, numerous biosensors coupled to microdialysis for the quantification of molecules found in body fluids, tissues, and organs other than the brain have been reported [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref] [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref] [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref] [bib_ref] Ex-Vivo Glucose Sensors Using Micro-Dialysis: Importance of On-Line Recovery Rate Determination by..., Vahlsing [/bib_ref] [bib_ref] Reliable Glucose Monitoring by Ex-Vivo Blood Microdialysis and Infrared Spectrometry for Patients..., Vahlsing [/bib_ref] [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref] [bib_ref] Continuous monitoring of adenosine and its metabolites using microdialysis coupled to microchip..., Gunawardhana [/bib_ref] [bib_ref] Bioelectronic organ-based sensor for microfluidic real-time analysis of the demand in insulin, Perrier [/bib_ref] [bib_ref] Extracellular hydrogen peroxide measurements using a flow injection system in combination with..., Moßhammer [/bib_ref] [bib_ref] PDMS/glass hybrid device with a reusable carbon electrode for on-line monitoring of..., Saylor [/bib_ref] [bib_ref] Continuous monitoring of bisulfide variation in microdialysis effluents by on-line droplet-based microfluidic..., Zhu [/bib_ref] [bib_ref] Microdialysis sampling coupled to microchip electrophoresis with integrated amperometric detection on an..., Scott [/bib_ref] [bib_ref] Microdialysis coupled with an embedded systems controller and CMOS image sensor, Rosenbloom [/bib_ref] [bib_ref] Randomized Trial of Closed-Loop Control in Very Young Children with Type 1..., Ware [/bib_ref] [bib_ref] 3D Printed Microfluidic Device with Integrated Biosensors for Online Analysis of Subcutaneous..., Gowers [/bib_ref]. An earlier review of biochemical monitoring of various bodily fluids has been presented elsewhere [bib_ref] Real-Time Clinical Monitoring of Biomolecules, Rogers [/bib_ref]. Examples of biosensors integrated with microdialysis for analyte detection are presented in [fig_ref] Figure 3: Biosensors coupled to microdialysis for quantification of clinical analytes [/fig_ref].
Neurochemical changes in TBI may also be propagated by an occurrence of spreading depolarisations, which originate from the lesion foci and spread out to neighbouring tissues at risk of secondary brain damage. This causes a dramatic disruption of ionic homeostasis, leading to a dramatic increase in extracellular potassium ions [bib_ref] Extracellular ion concentrations during spreading depression and ischemia in the rat brain..., Hansen [/bib_ref] [bib_ref] Mechanisms of spreading depression and hypoxic spreading depression-like depolarization, Somjen [/bib_ref] , neuronal and astrocytic swelling [bib_ref] Recurrent spontaneous spreading depolarizations facilitate acute dendritic injury in the ischemic penumbra, Risher [/bib_ref] [bib_ref] Persistent astroglial swelling accompanies rapid reversible dendritic injury during strokeinduced spreading depolarizations, Risher [/bib_ref] , glucose depletion and accumulation of lactate leading to tissue acidosis [bib_ref] Regulation of cerebral metabolism during cortical spreading depression, Feuerstein [/bib_ref] [bib_ref] Persisting depletion of brain glucose following cortical spreading depression, despite apparent hyperaemia:..., Hashemi [/bib_ref] [bib_ref] Continuous online microdialysis using microfluidic sensors: Dynamic neurometabolic changes during spreading depolarization, Rogers [/bib_ref] [bib_ref] Lactate and pH change in close correlation in the extracellular space of..., Scheller [/bib_ref] , and changes in cerebral blood flow [bib_ref] Spreading Depression, Spreading Depolarizations, and the Cerebral Vasculature, Ayata [/bib_ref]. The characteristic features of spreading depolarisations are a large transient negative shift in the slow electrical or direct current potential and the simultaneous silencing of brain electrical activity termed spreading depression [bib_ref] Mechanism of spreading cortical depression, Grafstein [/bib_ref] [bib_ref] Traumatic brain injury-induced autoregulatory dysfunction and spreading depression-related neurovascular uncoupling: Pathomechanisms, perspectives,..., Toth [/bib_ref]. This has led to the development of electrochemical-based potassium sensors integrated with microdialysis. An in-depth discussion of potassium biosensors and potassium measurements is beyond the scope of this review. However, their relevance to the field, especially when potassium measurements are interpreted along with neurochemical measurements, cannot be overlooked and have been briefly mentioned in Section 3. [fig_ref] Table 2: A summary comparison of recent key examples of biosensors integrated with microdialysis... [/fig_ref]. The use of various types of biosensors in clinical research is growing. [fig_ref] Figure 3: Biosensors coupled to microdialysis for quantification of clinical analytes [/fig_ref] highlights the diversity of the biosensor technologies and their respective setups in tandem with microdialysis. For instance, [fig_ref] Figure 3: Biosensors coupled to microdialysis for quantification of clinical analytes [/fig_ref] -D all have measured cerebral metabolites in dialysates of TBI patients, while [fig_ref] Figure 3: Biosensors coupled to microdialysis for quantification of clinical analytes [/fig_ref] has quantified blood in the dialysates of subcutaneous interstitial fluid [bib_ref] High-Performance Bioinstrumentation for Real-Time Neuroelectrochemical Traumatic Brain Injury Monitoring, Papadimitriou [/bib_ref] [bib_ref] Traumatic brain injury neuroelectrochemical monitoring: Behind-the-ear micro-instrument and cloud application, Tageldeen [/bib_ref] [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref]. The development of biosensors in monitoring cerebral metabolism is challenging, and this is evident as no biosensor has been approved for clinical use. The complexity and heterogeneity of TBI as well as the difficulty in fabricating ultrasensitive sensors for monitoring brain metabolism are factors slowing the progress. The development of a clinically effective biosensor requires optimisation and validation [bib_ref] Introduction to biosensors, Bhalla [/bib_ref]. For effective optimisation and validation, the sensor must satisfy several factors including selectivity, sensitivity, stability, reproducibility, linearity, and multiplexing [bib_ref] Introduction to biosensors, Bhalla [/bib_ref]. For ease and clear communication, each parameter has been defined in in the context of brain metabolism in TBI. The most common types of sensors developed for metabolite monitoring in TBI are electrochemical or optical, and they are thoroughly discussed in the next sections.
straightforward. However, based on a transducer, biosensors can be classified as electro-chemical, optical, and others [bib_ref] Introduction to biosensors, Bhalla [/bib_ref] [bib_ref] Recent advances of electrochemical and optical enzyme-free glucose sensors operating at physiological..., Adeel [/bib_ref] [bib_ref] Simultaneous optical and electrochemical label-free biosensing with ITO-coated lossy-mode resonance sensor, Śmietana [/bib_ref] [bib_ref] The role of biosensors in the detection of emerging infectious diseases, Pejcic [/bib_ref].
Ever since Clark and Lyons introduced a biosensor to monitor blood gas levels in 1962 [bib_ref] The future of biosensors, Bergveld [/bib_ref] , recent advances have facilitated alternative detection modalities. Since then, various types of biosensors for specific detection of analytes [bib_ref] Prostate cancer detection: A systematic review of urinary biosensors, Chan [/bib_ref] [bib_ref] Biosensors and nanotechnology for cancer diagnosis (lung and bronchus, breast, prostate, and..., Sharifianjazi [/bib_ref] have been developed; hence, this is now a multidisciplinary area bridging fundamental science (chemistry, physics, and biology) and clinical practice. An early example of a biosensor integrated with microdialysis for diagnostics was reported by Roda and colleagues in 1991 to monitor extracellular lactate in a rat brain [bib_ref] Microdialysis and luminescent probe: Analytical and clinical aspects, Roda [/bib_ref]. The enzyme-based electrochemical biosensor showed a high performance in quantifying lactate attributed to a limit of detection (LoD) of 0.1 mM, a wide linearity range between 0.1 and 100 mM, and an imprecision of less than 5%. Advantageously, this biosensor quantified other small molecules including glycerol. Since then, numerous biosensors coupled to microdialysis for the quantification of molecules found in body fluids, tissues, and organs other than the brain have been reported [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref] [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref] [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref] [bib_ref] Ex-Vivo Glucose Sensors Using Micro-Dialysis: Importance of On-Line Recovery Rate Determination by..., Vahlsing [/bib_ref] [bib_ref] Reliable Glucose Monitoring by Ex-Vivo Blood Microdialysis and Infrared Spectrometry for Patients..., Vahlsing [/bib_ref] [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref] [bib_ref] Continuous monitoring of adenosine and its metabolites using microdialysis coupled to microchip..., Gunawardhana [/bib_ref] [bib_ref] Bioelectronic organ-based sensor for microfluidic real-time analysis of the demand in insulin, Perrier [/bib_ref] [bib_ref] Extracellular hydrogen peroxide measurements using a flow injection system in combination with..., Moßhammer [/bib_ref] [bib_ref] PDMS/glass hybrid device with a reusable carbon electrode for on-line monitoring of..., Saylor [/bib_ref] [bib_ref] Continuous monitoring of bisulfide variation in microdialysis effluents by on-line droplet-based microfluidic..., Zhu [/bib_ref] [bib_ref] Microdialysis sampling coupled to microchip electrophoresis with integrated amperometric detection on an..., Scott [/bib_ref] [bib_ref] Microdialysis coupled with an embedded systems controller and CMOS image sensor, Rosenbloom [/bib_ref] [bib_ref] Randomized Trial of Closed-Loop Control in Very Young Children with Type 1..., Ware [/bib_ref] [bib_ref] 3D Printed Microfluidic Device with Integrated Biosensors for Online Analysis of Subcutaneous..., Gowers [/bib_ref]. An earlier review of biochemical monitoring of various bodily fluids has been presented elsewhere [bib_ref] Real-Time Clinical Monitoring of Biomolecules, Rogers [/bib_ref]. Examples of biosensors integrated with microdialysis for analyte detection are presented in [fig_ref] Figure 3: Biosensors coupled to microdialysis for quantification of clinical analytes [/fig_ref]. Proof-ofconcept design for a wireless, portable bio-instrument to monitor neurochemicals in brain injured patients [bib_ref] Traumatic brain injury neuroelectrochemical monitoring: Behind-the-ear micro-instrument and cloud application, Tageldeen [/bib_ref]. Previous work from the same group developed a biosensor setup (B): (I) is the combined needle with a PDMS microfluidic chip placed within a LabSmith automated platform for glucose detection, (II) tip of needle electrode, and (III) potassium ion selective electrode [bib_ref] High-Performance Bioinstrumentation for Real-Time Neuroelectrochemical Traumatic Brain Injury Monitoring, Papadimitriou [/bib_ref]. Similarly, (C) shows a setup as in (B) but includes a lactate biosensor and demonstrates integration with microdialysis in-vivo [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. (D) Illustration of an optical based sensor integrated with brain microdialysis for brain metabolite monitoring [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref]. (E) is another optical based biosensor set up for glucose monitoring in the dialysates of subcutaneous interstitial fluid [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref]. Images (A-D) originally appeared in [bib_ref] High-Performance Bioinstrumentation for Real-Time Neuroelectrochemical Traumatic Brain Injury Monitoring, Papadimitriou [/bib_ref] [bib_ref] Traumatic brain injury neuroelectrochemical monitoring: Behind-the-ear micro-instrument and cloud application, Tageldeen [/bib_ref] [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] Proof-ofconcept design for a wireless, portable bio-instrument to monitor neurochemicals in brain injured patients [bib_ref] Traumatic brain injury neuroelectrochemical monitoring: Behind-the-ear micro-instrument and cloud application, Tageldeen [/bib_ref]. Previous work from the same group developed a biosensor setup (B): (I) is the combined needle with a PDMS microfluidic chip placed within a LabSmith automated platform for glucose detection, (II) tip of needle electrode, and (III) potassium ion selective electrode [bib_ref] High-Performance Bioinstrumentation for Real-Time Neuroelectrochemical Traumatic Brain Injury Monitoring, Papadimitriou [/bib_ref]. Similarly, (C) shows a setup as in (B) but includes a lactate biosensor and demonstrates integration with microdialysis in-vivo [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. (D) Illustration of an optical based sensor integrated with brain microdialysis for brain metabolite monitoring [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref]. (E) is another optical based biosensor set up for glucose monitoring in the dialysates of subcutaneous interstitial fluid [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref]. Images (A-D) originally appeared in [bib_ref] High-Performance Bioinstrumentation for Real-Time Neuroelectrochemical Traumatic Brain Injury Monitoring, Papadimitriou [/bib_ref] [bib_ref] Traumatic brain injury neuroelectrochemical monitoring: Behind-the-ear micro-instrument and cloud application, Tageldeen [/bib_ref] [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. Important sensor parameters in the context of brain metabolism.
## Parameter meaning
## Selectivity
The sensor should be able to detect a molecule of interest, e.g., glucose, in the presence of other molecules and endogenous substances found in microdialysates.
## Sensitivity
The sensor should be able to detect the relevant range of metabolite concentrations seen in TBI.
## Stability
The sensor should not be influenced by changes in the external or internal environment when monitoring brain metabolism as this can lead to distortion of output signals.
## Linearity
The sensor should measure accurate concentration changes proportionately, and ideally in a linear manner.
## Reproducibility
The results acquired by the sensor should be of highest accuracy and the investigator should be confident of achieving the same results if the brain metabolic conditions were constant.
## Multiplexing
Ability of the sensor to detect several analytes simultaneously, e.g., glucose, lactate, and pyruvate, in microdialysates containing these and other endogenous molecules.
## Electrochemical sensors
Electrochemical sensors are widely used for the detection of metabolites owing to their fast, precise, selective, and easy-to-use capabilities. They operate by producing an electrical signal proportional to the concentration of the analyte of interest [bib_ref] Introduction to biosensors, Bhalla [/bib_ref]. Generally, chemical sensors rely on conductometric, potentiometric, or amperometric measurements [bib_ref] Chapter One-Advances in electrochemical immunosensors, Aydin [/bib_ref].
In principle, most commonly used biosensors utilise a three-electrode systemconsisting of a working electrode (WE), a reference electrode (RE), and a counter electrode (CE).schematically shows how enzymes play an important role in electrochemical detection, whereby an electrical current arising during oxidation (reduction) of glucose on the surface of a working electrode, under certain potentials, is measured. Electrochemical methods for analysing glucose typically use the enzyme glucose oxidase (abbreviated as GOx or GOD in the literature) to catalyse the conversion of glucose to gluconolactone. In some assay formats, the oxidising agent is molecular oxygen (O 2 ), which thereby becomes chemically reduced to form hydrogen peroxide that is detected with an electrode (Equation (3)) [bib_ref] Ferrocenemediated enzyme electrode for amperometric determination of glucose, Cass [/bib_ref].
[formula] Glucose + O 2 → gluconolactone + H 2 O 2(3) [/formula]
However, this measurement can be unstable if O 2 levels fluctuate for any reason in the sample, e.g., in blood or tissue, or if another redox active species such as ascorbate is present. Therefore, some methods use another molecule as an artificial electron charge transfer moiety "mediator" [bib_ref] 5-Practical applications and protocols for enzyme biosensors, Davis [/bib_ref] , e.g., a ferrocene derivative (Fc), that acts preferentially as an oxidant with the glucose oxidase enzyme to "shuttle" electrons between the enzyme and the electrode-see Equation (4) [bib_ref] Ferrocenemediated enzyme electrode for amperometric determination of glucose, Cass [/bib_ref]. This lowers the electrical potential for measuring the enzyme-catalysed reaction, thus diminishing the interference by redox active species present within the sample.
[formula] Glucose + GOx (ox) → gluconolactone + GOx (red)(4) [/formula]
GOx (red) + 2Fc + → GOx (ox) + 2Fc + 2H + 2Fc 2Fc + + 2e − 022, 12, x FOR PEER REVIEW However, this measurement can be unstable if the sample, e.g., in blood or tissue, or if another red present. Therefore, some methods use another mo transfer moiety "mediator" [bib_ref] 5-Practical applications and protocols for enzyme biosensors, Davis [/bib_ref] , e.g., a ferrocene der an oxidant with the glucose oxidase enzyme to "sh and the electrode-see Equation (4) [bib_ref] Ferrocenemediated enzyme electrode for amperometric determination of glucose, Cass [/bib_ref]. This lowers the enzyme-catalysed reaction, thus diminishing th present within the sample. In addition to TBI, it is worth mentioning the usefulness of combining microdialysis with biosensors in other disorders including diabetes mellitus and neurodegeneration. To better understand the neurochemistry of neurodegenerative diseases, Gunawardhana and Lunte have developed a separation-based sensor using microchip electrophoresis with electrochemical detection integrated with microdialysis for continuous measurements of adenosine and subsequent downstream metabolites [bib_ref] Continuous monitoring of adenosine and its metabolites using microdialysis coupled to microchip..., Gunawardhana [/bib_ref]. Perrier and colleagues have also developed a microfluidic platinum black microelectrode array with microdialysis integration capabilities for real-time glucose monitoring in pancreatic islets [bib_ref] Bioelectronic organ-based sensor for microfluidic real-time analysis of the demand in insulin, Perrier [/bib_ref].
## + →
In TBI, most electrochemical sensors for neurochemical measurements are based on amperometric detection using specific enzymes that are selective to relevant brain metabolites. Papadimitriou and colleagues have developed an enzymatic-based sensor integrated with microdialysis for the amperometric detection of glucose [bib_ref] High-Performance Bioinstrumentation for Real-Time Neuroelectrochemical Traumatic Brain Injury Monitoring, Papadimitriou [/bib_ref]. Here, Papadimitriou et al. used glucose oxidase (Gox) and horseradish peroxidase (HRP) enzymes in 1.5 mM ferrocene monocarboxylic acid (Fc) solution that is immobilised onto a combined needle electrode within a microfluidic channel. The glucose-Gox reaction generates hydrogen peroxide that reacts with the mixture of HRP and Fc to form ferrocenium ions. This reduction process is quantifiable at the surface electrode, allowing the acquisition of glucose measurements. The system's setup is displayed in [fig_ref] Figure 3: Biosensors coupled to microdialysis for quantification of clinical analytes [/fig_ref] , while technical details are communicated in. Another enzymatic-based biosensor integrated with a microdialysis stream using a microfluidic PDMS chip by Pagkalos and colleagues has measured lactate concentration changes in vitro [bib_ref] A Higà Performance Application Specific Integrated Circuit for Electrical and Neurochemical Traumatic..., Pagkalos [/bib_ref]. Here, they detected lactate levels between 0 and 50 µM with stepwise changes of 12.5 µM every 4 min. For continuous measurements, Robbins et al. reported below 10 µM extracellular glucose via dexamethasone-enhanced microdialysis coupled with an enzymatic-based sensor in ten rats [bib_ref] Extended (10-Day) Real-Time Monitoring by Dexamethasone-Enhanced Microdialysis in the Injured Rat Cortex, Robbins [/bib_ref]. Other examples of neurochemical measurements using electrochemical sensors have been presented in [fig_ref] Table 2: A summary comparison of recent key examples of biosensors integrated with microdialysis... [/fig_ref] and [fig_ref] Figure 5: Brain metabolic changes measured by electrochemical biosensors [/fig_ref]. microdialysis stream using a microfluidic PDMS chip by Pagkalos and colleagues has measured lactate concentration changes in vitro [bib_ref] A Higà Performance Application Specific Integrated Circuit for Electrical and Neurochemical Traumatic..., Pagkalos [/bib_ref]. Here, they detected lactate levels between 0 and 50 μM with stepwise changes of 12.5 μM every 4 min. For continuous measurements, Robbins et al. reported below 10 μM extracellular glucose via dexamethasone-enhanced microdialysis coupled with an enzymatic-based sensor in ten rats [bib_ref] Extended (10-Day) Real-Time Monitoring by Dexamethasone-Enhanced Microdialysis in the Injured Rat Cortex, Robbins [/bib_ref].
Other examples of neurochemical measurements using electrochemical sensors have been presented in [fig_ref] Table 2: A summary comparison of recent key examples of biosensors integrated with microdialysis... [/fig_ref] and [fig_ref] Figure 5: Brain metabolic changes measured by electrochemical biosensors [/fig_ref]. [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. (B) Microdialysate measurements affected by an air bubble [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref]. (C) Simultaneous glucose and potassium changes over time in a patient 6 days after a TBI. The dotted circles indicate potassium transients while the red allows indicates glucose decline in relation to potassium rise (purple arrow) [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref]. Images (A-C) originally appeared in [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] , [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref] and [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref] respectively, published Open Access CC BY [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref] and CC BY-NC-ND [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref] , copyright The Authors.
A few studies have shown both glucose and lactate measurements in TBI using electrochemical sensors integrated with microdialysis. Tageldeen and colleagues have achieved both glucose and lactate measurements using wearable electrochemical sensors [bib_ref] Traumatic brain injury neuroelectrochemical monitoring: Behind-the-ear micro-instrument and cloud application, Tageldeen [/bib_ref]. They were able to demonstrate dynamic metabolic changes ranging from 0 to 1 mM with 0.25 mM steps. Glucose and lactate calibration curves were produced using linear regression of R 2 = 0.998 and R 2 = 0.995, respectively, and the LoDs for glucose and lactate were 0.85 μM and 1.2 μM, respectively.
Although in vitro studies enhance our understanding of neurochemical monitoring in TBI, their limitation in recapitulating the brain's complex and dynamic changes means that simple non-biological model findings may not be replicated in humans. Continuously recorded microdialysis data in six patients for a period of over 6 h has shown that changes in the levels of glucose and lactate influence spreading depolarisations, as evidenced by increases in potassium levels [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. In one of the patients [fig_ref] Figure 5: Brain metabolic changes measured by electrochemical biosensors [/fig_ref] , they found basal glucose levels in the microdialysate to be low (100 μM) despite high blood glucose levels [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. (B) Microdialysate measurements affected by an air bubble [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref]. (C) Simultaneous glucose and potassium changes over time in a patient 6 days after a TBI. The dotted circles indicate potassium transients while the red allows indicates glucose decline in relation to potassium rise (purple arrow) [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref]. Images (A-C) originally appeared in [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] , [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref] and [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref] respectively, published Open Access CC BY [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref] and CC BY-NC-ND [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref] , copyright The Authors.
A few studies have shown both glucose and lactate measurements in TBI using electrochemical sensors integrated with microdialysis. Tageldeen and colleagues have achieved both glucose and lactate measurements using wearable electrochemical sensors [bib_ref] Traumatic brain injury neuroelectrochemical monitoring: Behind-the-ear micro-instrument and cloud application, Tageldeen [/bib_ref]. They were able to demonstrate dynamic metabolic changes ranging from 0 to 1 mM with 0.25 mM steps. Glucose and lactate calibration curves were produced using linear regression of R 2 = 0.998 and R 2 = 0.995, respectively, and the LoDs for glucose and lactate were 0.85 µM and 1.2 µM, respectively.
Although in vitro studies enhance our understanding of neurochemical monitoring in TBI, their limitation in recapitulating the brain's complex and dynamic changes means that simple non-biological model findings may not be replicated in humans. Continuously recorded microdialysis data in six patients for a period of over 6 h has shown that changes in the levels of glucose and lactate influence spreading depolarisations, as evidenced by increases in potassium levels [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. In one of the patients [fig_ref] Figure 5: Brain metabolic changes measured by electrochemical biosensors [/fig_ref] , they found basal glucose levels in the microdialysate to be low (100 µM) despite high blood glucose levels (11.8 mM) reflecting the brain tissue's high energy demand, while a high lactate concentration is attributed to anaerobic glycolytic metabolism in the injured brain [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref]. In another study, Gowers and colleagues also continuously monitored cerebral metabolites found in dialysates [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref]. Here, they first tested their system in an in vitro brain model before testing it on patients. For non-clinical data, the sensor was continuously calibrated using physiological buffer and 2 mM glucose concentration with steps of 0.5 mM at a constant flow rate of 0.2 µL/min. The quality of the data acquired in a clinical setting was hindered by the air bubbles in the system leading to decreased signal-to-noise ratio [fig_ref] Figure 5: Brain metabolic changes measured by electrochemical biosensors [/fig_ref]. However, following three hours of monitoring, the sensor detected a sudden lactate increase in one patient. Vital information such as this has also been presented by in a follow-up study to the one previously discussed [bib_ref] Extended (10-Day) Real-Time Monitoring by Dexamethasone-Enhanced Microdialysis in the Injured Rat Cortex, Robbins [/bib_ref]. Here, Gifford and colleagues reported declining glucose levels in three TBI patients and persistent low glucose in one TBI patient. In one of the patients, glucose levels declined to non-detectable levels, as evidenced in [fig_ref] Figure 5: Brain metabolic changes measured by electrochemical biosensors [/fig_ref]. This long-term glucose decline has been said to be because of metabolic aberrations in an injured brain [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref].
## Optical biosensors
Optical biosensors offer many advantages over conventional analytical methods owing to their direct, real-time, and label-free detection of many biochemical substrates. Optical detection works by exploiting the interaction of light with the analyte of interest or with a recognition element that is selective to the analyte. Various applications of optical sensing have been performed both in a label-free or label-based manner. For label-based detection, the optical signal is acquired by fluorescent or colorimetric techniques [bib_ref] Highly specific C-C bond cleavage induced FRET fluorescence for in vivo biological..., Li [/bib_ref] [bib_ref] Detection of Brain Tumors and Systemic Metastases Using NanoLuc and Fluc for..., Germain-Genevois [/bib_ref] [bib_ref] Enhanced detection of myeloperoxidase activity in deep tissues through luminescent excitation of..., Zhang [/bib_ref] [bib_ref] Nanoarchitecture Based SERS for Biomolecular Fingerprinting and Label-Free Disease Markers Diagnosis, Sinha [/bib_ref] , whereas label-free techniques involve light-analyte interactions to facilitate signal detection exemplified by spectroscopic biosensors [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref] [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref] [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref] [bib_ref] Ex-Vivo Glucose Sensors Using Micro-Dialysis: Importance of On-Line Recovery Rate Determination by..., Vahlsing [/bib_ref] [bib_ref] Reliable Glucose Monitoring by Ex-Vivo Blood Microdialysis and Infrared Spectrometry for Patients..., Vahlsing [/bib_ref] [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref] [bib_ref] Evaluation and Benchmarking of an EC-QCL-Based Mid-Infrared Spectrometer for Monitoring Metabolic Blood..., Grafen [/bib_ref]. Other than the widely used offline ISCUSflex colorimetric bedside analyser, all optical based techniques integrated with microdialysis to quantify TBI metabolites have utilised infrared spectroscopic sensors.
Infrared spectroscopy has been successfully used for numerous applications in biochemistry as evidenced in the literature offering innovative diagnostic methods [bib_ref] Evaluation and Benchmarking of an EC-QCL-Based Mid-Infrared Spectrometer for Monitoring Metabolic Blood..., Grafen [/bib_ref] [bib_ref] Simultaneous determination of glucose, triglycerides, urea, cholesterol, albumin and total protein in..., Jessen [/bib_ref] [bib_ref] Rapid Label-Free Analysis of Brain Tumor Biopsies by Near Infrared Raman and..., Galli [/bib_ref]. Specifically, mid-infrared (MIR) regions ranging from 2.5 to 25 µm has been primarily adopted for monitoring neurochemicals. Furthermore, investigators have opted for powerful mid-infrared light sources such as quantum cascade lasers (QCLs) for spectrometric analysis [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref] [bib_ref] Evaluation and Benchmarking of an EC-QCL-Based Mid-Infrared Spectrometer for Monitoring Metabolic Blood..., Grafen [/bib_ref].
Infrared spectroscopy allows the investigation of the interaction of light with organic compounds via the excitation of vibrational and rotational modes. The principle of measurement is shown in [fig_ref] Figure 6: Basic principles of infrared spectroscopy [/fig_ref] , where infrared light passes through a sample containing the analytes of interest and MIR radiation interacts with microdialysate samples, providing a spectral fingerprint that is useful for identifying the molecules of interest such as glucose, lactate, and pyruvate. Supporting the absorbance analysis theory is the Beer-Lambert law [bib_ref] 2-PHOTOPHYSICS, Wypych [/bib_ref] depicted in Equation (5).
[formula] A = log 1 T = εLc(5) [/formula]
Here, A is absorbance, T is transmittance, ε is the molar absorptivity coefficient in L/((mol)(cm)) of the metabolite of interest at a certain wavelength, L is the path-length in cm through solution, and c is the concentration in molarity (M) of the metabolite of interest. Glucose, lactate, and pyruvate have varying absorbance intensities at specific wavelengths. The degree of absorption is determined by the wavelength of light, the absorption cross-section of the analyte, and the optical path-length. compounds via the excitation of vibrational and rotational modes. The principle of meas urement is shown in [fig_ref] Figure 6: Basic principles of infrared spectroscopy [/fig_ref] , where infrared light passes through a sample containing the analytes of interest and MIR radiation interacts with microdialysate samples, provid ing a spectral fingerprint that is useful for identifying the molecules of interest such a glucose, lactate, and pyruvate. Supporting the absorbance analysis theory is the Beer-Lambert law [bib_ref] 2-PHOTOPHYSICS, Wypych [/bib_ref] depicted in Equation (5). The versatility of optical sensing technologies integrated with microdialysis for reagent-free multi-analyte assays has been demonstrated by various studies [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref] [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref] [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref] [bib_ref] Ex-Vivo Glucose Sensors Using Micro-Dialysis: Importance of On-Line Recovery Rate Determination by..., Vahlsing [/bib_ref] [bib_ref] Reliable Glucose Monitoring by Ex-Vivo Blood Microdialysis and Infrared Spectrometry for Patients..., Vahlsing [/bib_ref] [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref] [bib_ref] Evaluation and Benchmarking of an EC-QCL-Based Mid-Infrared Spectrometer for Monitoring Metabolic Blood..., Grafen [/bib_ref]. Examples are presented in [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref]. Heise and colleagues have developed an optical sensing system integrated with microdialysis for continuous measurements of analytes found in blood [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref] [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref]. Using infrared transmission-mode, the system measured varying glucose concentrations between 5.2 and 15.8 mM for more than 90 h. Subsequently, they determined carbon dioxide and bicarbonate levels in blood to examine acid-base status of the body [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref]. Other than ex vivo analysis of glucose levels in heparinised blood of type 1 diabetic patients [bib_ref] Reliable Glucose Monitoring by Ex-Vivo Blood Microdialysis and Infrared Spectrometry for Patients..., Vahlsing [/bib_ref] , Vahlsing and colleagues have also simultaneously quantified glucose and mannitol in recovery-rate studies, as well as lactate, demonstrating the capabilities of the optical sensor to quantify relevant clinical substrates [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref].
The developed infrared spectrometric sensor by Vahlsing and colleagues consists of a 30-micrometre pathlength transmission micro-cell with custom-built microfluidics that permits continuous automated sample measurements [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref]. highlights the correlation and accuracy between a commercially available Beckman glucose analyser and their novel optical sensor in measuring blood glucose. Subsequently, lactate concentrations determined from the microdialysates spectra were compared against the reference lactate assay from 5 µL volumes of intermittently acquired dialysates. As shown in , as the lactate concentration normalises over time, the correlation of interstitial glucose to blood glucose is established [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref].
As previously mentioned, advances in the field have led to the use of powerful light-sources to enhance accuracy, precision, and long-term stability when measuring relevant clinical metabolites [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Evaluation and Benchmarking of an EC-QCL-Based Mid-Infrared Spectrometer for Monitoring Metabolic Blood..., Grafen [/bib_ref]. Evidence shows that the use of powerful light sources, such as QCLs, instead of thermal emitters to quantify metabolites, yields better overall results [bib_ref] Evaluation and Benchmarking of an EC-QCL-Based Mid-Infrared Spectrometer for Monitoring Metabolic Blood..., Grafen [/bib_ref]. Using a QCL-based spectrometric sensor integrated with microdialysis, Alimagham and colleagues quantified dialysates from two traumatically brain-injured patients acquired over the course of 48 h [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref]. The analysis demonstrated the ability of the novel biosensor in detecting and measuring glucose and lactate in microdialysates of TBI patients. The biosensor exhibits LoDs of 0.5, 0.2, and 0.1 mM for glucose, lactate, and pyruvate, respectively. In vitro measurements of glucose, lactate, and pyruvate showed good linearity of the sensor, as shown in -E, respectively. Furthermore, highlighting the accuracy of the technology was the correlation of the results acquired using the QCL-based spectrometric sensor and the well-established enzymatic colorimetric assays on the ISCUSflex microdialysis analyser whereby glucose, lactate, and pyruvate levels were measured from microdialysates of two TBI patients ex vivo .
The versatility of optical sensing technologies integrated with microdialysis for reagent-free multi-analyte assays has been demonstrated by various studies [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref] [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref] [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref] [bib_ref] Ex-Vivo Glucose Sensors Using Micro-Dialysis: Importance of On-Line Recovery Rate Determination by..., Vahlsing [/bib_ref] [bib_ref] Reliable Glucose Monitoring by Ex-Vivo Blood Microdialysis and Infrared Spectrometry for Patients..., Vahlsing [/bib_ref] [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref] [bib_ref] Evaluation and Benchmarking of an EC-QCL-Based Mid-Infrared Spectrometer for Monitoring Metabolic Blood..., Grafen [/bib_ref]. Examples are presented in [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref]. Heise and colleagues have developed an optical sensing system integrated with microdialysis for continuous measurements of analytes found in blood [bib_ref] Bedside monitoring of subcutaneous interstitial glucose in healthy individuals using microdialysis and..., Heise [/bib_ref] [bib_ref] Microdialysis Based Monitoring of Subcutaneous Interstitial and Venous Blood Glucose in Type..., Heise [/bib_ref]. Using infrared transmission-mode, the system measured varying glucose concentrations between 5.2 and 15.8 mM for more than 90 h. Subsequently, they determined carbon dioxide and bicarbonate levels in blood to examine acid-base status of the body [bib_ref] Monitoring of interstitial buffer systems using micro-dialysis and infrared spectrometry, Heise [/bib_ref]. Other than ex vivo analysis of glucose levels in heparinised blood of type 1 diabetic patients [bib_ref] Reliable Glucose Monitoring by Ex-Vivo Blood Microdialysis and Infrared Spectrometry for Patients..., Vahlsing [/bib_ref] , Vahlsing and colleagues have also simultaneously quantified glucose and mannitol in recovery-rate studies, as well as lactate, demonstrating the capabilities of the optical sensor to quantify relevant clinical substrates [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref]. [bib_ref] Combination of micro-dialysis and infrared spectroscopy: A multianalyte assay for accurate biofluid..., Vahlsing [/bib_ref]. Images (C-G) originally appeared in [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref] , published Open Access CC BY, copyright The Authors.
## Progress, challenges, and future perspectives
The sensing techniques reviewed here have shown promising results in providing metabolic insights of the injured brain. The enzymatic colorimetric method is used in the standard analyser for the quantification of the relevant TBI metabolites. High accuracy and precision in quantifying metabolic values including LPR justifies why it is the most reliable technique to date. However, hourly non-continuous measurements and the need for reagents are the main obstacles. Optical infrared spectroscopic sensors offer quick, precise, and reagent-free analysis, and this technique has shown promise in quantifying the brain metabolites in TBI. However, MIR technology's feasibility to continuously monitor cerebral metabolites reliably online needs more evidence. Although electrochemical biosensors exhibit good analytical performance, their sensitivities gradually decrease with time when in use. Overall, this exhibits a challenge in recording the fluctuating metabolic concentrations often seen in severe TBI patients. Therefore, counteracting the gradual loss of sensitivities of these biosensors would facilitate high accuracy and reliable data acquired when monitoring the traumatically injured brain. M Dialysis AB developed LOKE, a portable commercially available enzymaticelectrochemical sensor for continuous online clinical microdialysis monitoring. This automated biosensor relies on enzymatic detection to generate absolute glucose and lactate values, plus trends, and can run for up to 5 days. The advantage for busy clinical staff is that LOKE eliminates the need for manual hourly assays on ISCUSflex. However, the detection of pyruvate is work in progress. Once all three relevant brain metabolites become quantifiable, this would be a remarkable advance in the field. A limitation of LOKE is that it should only be used with microdialysis catheters with 20 kDa membrane pore size. Boutelle's group has already continuously monitored microdialysates online using home-built enzymatic-electrochemical-based sensors, and some of their findings have elucidated the pathophysiological changes in TBI patients, highlighting the utility of integrating biosensors with microdialysis [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref] [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref]. However, some data acquired had noise and artefacts that contributed to the loss of vital information [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref]. All biosensors mentioned in [fig_ref] Table 2: A summary comparison of recent key examples of biosensors integrated with microdialysis... [/fig_ref] are in the design and validation phase, with only work in references [bib_ref] Simultaneous monitoring of potassium, glucose and lactate during spreading depolarization in the..., Rogers [/bib_ref] [bib_ref] Validation of Dexamethasone-Enhanced Continuous-Online Microdialysis for Monitoring Glucose for 10 Days after..., Gifford [/bib_ref] [bib_ref] Clinical translation of microfluidic sensor devices: Focus on calibration and analytical robustness, Gowers [/bib_ref] , testing their technologies on patients. Thus, there is a need for more validation studies in humans to establish their efficacy and robustness. Subsequently, these biosensors could be optimised for enhanced performance to achieve real-time continuous online microdialysis monitoring in TBI patients. Furthermore, implantable non-microdialysis sensors for direct monitoring of neurochemicals in the brain offer improved temporal resolution [bib_ref] Fiber-Based Electrochemical Biosensors for Monitoring pH and Transient Neurometabolic Lactate, Booth [/bib_ref]. Booth and colleagues have measured lactate in the frontal and parietal cortex of mice using a fibre-based enzymatic sensor in situ [bib_ref] Fiber-Based Electrochemical Biosensors for Monitoring pH and Transient Neurometabolic Lactate, Booth [/bib_ref]. Although in its infancy, the development of specific, ultra-sensitive, and stable probes for in vivo direct monitoring of brain metabolism will advance real-time brain monitoring of TBI patients. Therefore, more studies to further improve emerging techniques for clinical use are needed.
Future studies should also focus on the development of data ecosystems encompassing brain metabolism and other monitored parameters such as ICP and PBtO 2 due to an interplay between cerebral metabolic demand and energy substrate delivery. For instance, brain oedema following TBI leads to an increase in ICP, which then causes a decrease in CPP, cerebral blood flow, and brain tissue oxygenation [bib_ref] Multimodality neuromonitoring in severe pediatric traumatic brain injury, Young [/bib_ref]. Due to this mismatch between cerebral metabolic demand and energy substrate delivery, the interpretation of brain chemistry should be in the context of the clinical condition of the patient and in conjunction with other monitored parameters including ICP, CPP, PBtO 2 , and cerebral vascular pressure reactivity (PRx) [bib_ref] Matrix Metalloproteinase Expression in Contusional Traumatic Brain Injury: A Paired Microdialysis Study, Guilfoyle [/bib_ref] [bib_ref] Brain tissue oxygen tension monitoring in pediatric severe traumatic brain injury. Part..., Figaji [/bib_ref]. Thus, data acquired using biosensor technologies when incorporated with data from other monitored parameters can improve methods of monitoring real-time cerebral physiology to better understand secondary brain injury onset and treatment strategies. When used as a guide for medical intervention , measurements acquired using these biosensors can aid in the prognostication of patient outcome. In addition, clinicians working in neuro-intensive care will benefit from improved clinical workflow facilitated by data ecosystems, allowing them to focus more of their time on direct patient care.
# Conclusions
In this review, we have summarised biosensors coupled to microdialysis for the detection of cerebral metabolites in TBI. To date, electrochemical and optical sensors are the only types of biosensors used for cerebral metabolite measurements. The accuracy and actionable information provided by these emerging technologies can help inform clinical decision making as well as aid in the prognostication of patient outcome. Other than improved performances, feasibility and robustness of these biosensors can be enhanced by conducting in vivo clinical studies. Future designs of the biosensor technology will include ultra-sensitive and stable probes for in situ direct sensing of brain metabolites for true real-time brain monitoring in TBI patients. [bib_ref] Cerebral Microdialysate Metabolite Monitoring using Mid-infrared Spectroscopy, Alimagham [/bib_ref]. The authors declare no other competing financial interest.
## Abbreviations
[fig] Figure 2: Cerebral microdialysis. (A) Triple lumen cranial access device (CAD) that allows micro alysis (MD), intracranial pressure (ICP) [/fig]
[fig] Figure 3: Biosensors coupled to microdialysis for quantification of clinical analytes. (A) [/fig]
[fig] Figure 4: (A) Schematic diagram of a three-electrode syst schematic presented here for clarity is also a representatio of glucose at an electrode mediated by a ferrocene deriva zyme that catalyses glucose oxidation. [/fig]
[fig] Figure 5: Brain metabolic changes measured by electrochemical biosensors. (A) Potassium, glucose and lactate changes over time in a TBI patient [/fig]
[fig] Figure 6: Basic principles of infrared spectroscopy. (1) Incident broadband infrared light (I0) upon (2) sample leads to vibrational modes of a molecule. Consequently, specific amounts of energy from the incident light are absorbed. This decreases the subsequently measured infrared light (I) by th [/fig]
[fig] Figure 7, Figure 7: Measurements of various metabolites using optical biosensors. Glucose (A) and lactate (B) continuous measurements of concentration versus time, determined in a type 1 diabetic subject. The inset graph within (A) shows % recovery versus time [69]. The arrows within (A) indicate the time between which the tissue lactate concentrations in (B) were also followed. (C) Glucose, (D) lactate, Measurements of various metabolites using optical biosensors. Glucose (A) and lactate (B) continuous measurements of concentration versus time, determined in a type 1 diabetic subject. The inset graph within (A) shows % recovery versus time [69]. The arrows within (A) indicate the time between which the tissue lactate concentrations in (B) were also followed. (C) Glucose, (D) lactate, and (E) pyruvate MIR measurements of absorbance vs. wavenumber (cm −1 ) for different concentrations of pure standards, whereby the insets are standard calibration curves of three replicates at specific wavelengths of each metabolite. (F,G) compare consecutively pooled microdialysate measurements of one TBI patient acquired by an optical based sensor to routinely used ISCUSflex microdialysis analyser where glucose and lactate was detected. The diagrams in (A,B) are published with permission of the Publisher and Corresponding Author*; they originally appeared in T. Vahlsing, S. Delbeck, J. Budde, D. Ihrig, and H. M. Heise*. Combination of microdialysis and infrared spectroscopy: a multi-analyte assay for accurate biofluid analysis and patient monitoring. Biomedical Vibrational Spectroscopy 2016: Advances in Research and Industry, edited by A. Mahadevan-Jansen, W. Petrich, Proc. of SPIE Vol. 9704, 97040R. © 2016 SPIE·CCC code: 1605-7422/16/$18, doi: 10.1117/12.2214636 [/fig]
[fig] Funding: We gratefully acknowledge support from the following: National Institute for Health Research Invention for Innovation Awards (NIHR i4i Challenge Award II-C5-0715-20005 and NIHR i4i Product Development Award NIHR200986) and the NIHR Brain Injury MedTech Co-operative. The APC was funded by NIHR i4i Product Development Award NIHR200986. C.Z. is supported by a Research Studentship from the W.D. Armstrong Trust (University of Cambridge). F.C.A. is supported by an NIHR i4i Product Development Award. K.L.H.C. is supported by the NIHR Biomedical Research Centre, Cambridge, and by an NIHR i4i Product Development Award. P.J.H. is supported by the NIHR Biomedical Research Centre, Cambridge, NIHR Senior Investigator Award, and the Royal College of Surgeons of England. The views expressed are those of the authors and are not necessarily those of the NIHR or the Department of Health and Social Care. Conflicts of Interest: The authors declare the following competing financial interest: F.C.A., K.L.H.C., P.J.H., and T.H. are inventors on a patent application related to the cited research in [/fig]
[table] Table 2: A summary comparison of recent key examples of biosensors integrated with microdialysis for brain metabolite monitoring[45][46][47][48][49][50][51][52]. [/table]
|
Cardiovascular Disease Performance Measures in the Outpatient Setting in India: Insights From the American College of Cardiology’s PINNACLE India Quality Improvement Program (PIQIP)
Background--India has a growing burden of cardiovascular disease (CVD), yet data on the quality of outpatient care for patients with coronary artery disease, heart failure, and atrial fibrillation in India are very limited. We collected data on performance measures for 68 196 unique patients from 10 Indian cardiology outpatient departments from January 1, 2011, to February 5, 2014, in the American College of Cardiology's PINNACLE (Practice Innovation and Clinical Excellence) India Quality Improvement Program (PIQIP). PIQIP is India's first national outpatient CVD quality-improvement program.Methods and Results--In the PIQIP registry, we estimated the prevalence of CVD risk factors (hypertension, diabetes, dyslipidemia, and current tobacco use) and CVD among outpatients. We examined adherence with performance measures established Physician Consortium for Performance Improvement for coronary artery disease, heart failure, and atrial fibrillation. There were a total of 68 196 patients (155 953 patient encounters), with a mean age of 50.6 years (SD 18.2 years). Hypertension was present in 29.7% of patients, followed by diabetes (14.9%), current tobacco use (7.6%), and dyslipidemia (6.5%). Coronary artery disease was present in 14.8%, heart failure was noted in 4.0%, and atrial fibrillation was present in 0.5% of patients. Among eligible patients, the reported use of medications was as follows: aspirin in 48.6%, clopidogrel in 37.1%, and statin-based lipidlowering therapy in 50.6% of patients with coronary artery disease; RAAS (renin-angiotensin-aldosterone system) antagonist in 61.9% and beta-blockers in 58.1% of patients with heart failure; and oral anticoagulants in 37.0% of patients with atrial fibrillation.Conclusions--This pilot study, initiated to improve outpatient CVD care in India, presents our preliminary results and barriers to data collection and demonstrates that such an initiative is feasible in a resource-limited environment. In addition, we attempted to outline areas for further improvement in outpatient CVD care delivery in India. ( J Am Heart
I ndia is currently the second most populous country in the world, with a population of >1.2 billion.Although there is a paucity of contemporary data on the prevalence of coronary artery disease (CAD) in India, various preliminary studies have reported high CAD prevalence, approaching 11% in the urban population and 7% in the rural population. [bib_ref] A survey for prevalence of coronary artery disease and its risk factors..., Murthy [/bib_ref] It has been estimated that 9.2 million productive years of life in India were lost to cardiovascular disease (CVD) in the year 2000, a number that is expected to rise to 18 million by 2030 (10 times the burden in the United States). [bib_ref] The burden of cardiovascular disease in the Indian subcontinent, Goyal [/bib_ref] There has also been an increase in the prevalence of traditional CAD risk factors in India because of a combination of behavioral risk factors, such as tobacco use, physical inactivity, and diets low in fruit and vegetable content, paralleled by low levels of awareness and treatment and control of CAD risk factors. [bib_ref] Risk factor profile for chronic non-communicable diseases: results of a community-based study..., Thankappan [/bib_ref] Although prior facility-based research on cardiovascular care in India has focused on in-hospital processes and outcome measures for stroke and myocardial infarction, for example, through the Kerala Acute Coronary Syndrome Registry, [bib_ref] Mohanan PP; Kerala Acute Coronary Syndrome Registry Investigators. Optimal in-hospital and discharge..., Huffman [/bib_ref] data on outpatient facility-based cardiovascular care in India are limited, with no widespread mechanism to offer feedback to clinicians on their performance with outpatient CVD care. Furthermore, there are only 4000 cardiologists in India, and the ratio of physicians to cardiac patients is disproportionately low.Popular medical care providers may see >200 patients each day. [bib_ref] The quality of medical advice in low-income countries, Das [/bib_ref] Consequently, busy practitioners are too overwhelmed to report or attempt to improve the quality of outpatient CVD care.Accordingly, the American College of Cardiology (ACC), in collaboration with local Indian cardiologists interested in participating in the Practice Innovation and Clinical Excellence (PINNACLE) India Quality Improvement Program (PIQIP), began data collection on the quality of outpatient care for CVD risk factors (hypertension, diabetes mellitus, current tobacco use) and disease states (CAD, heart failure [HF], atrial fibrillation [AF]) in India for the PIQIP, an extension of the ACC's US-based PINNACLE registry. [bib_ref] Cardiac performance measure compliance in outpatients: the American College of Cardiology and..., Chan [/bib_ref] The PINNACLE registry is a national prospective office-based quality-improvement program for cardiac care, designed to capture, report, and improve compliance with outpatient performance measures in the United States. 8
# Methods
## Data collection in piqip
We extended the PINNACLE registry to India in 2011 under PIQIP, the first outpatient CVD care data collection and reporting program implemented in the country [fig_ref] Figure 1: Infographic demonstrating, in a chronological fashion, the data-collection steps that were implemented... [/fig_ref]. Prior to its inception, local Indian cardiologists were contacted for evaluation of the feasibility of data-collection methods (eg, paper, Web-based, tablet, and local versus cloud-based storage) and for modification of the PINNACLE data dictionary and paper data-collection form. Following establishment of data-collection methods and modification of the data-collection form to meet the requirements of the Indian cardiologists, a small-scale pilot project was conducted, including 300 unique patient visits and 6 cardiologists from the states of Bihar and Maharashtra, to assess data-capture feasibility. The PIQIP Patient Care Tool (https://ehr.pinnaclehqi.org/PIQ-IPHQI/login.aspx), a Web-based data-collection form, was also designed based on a 1-month field trial at Ruby Hospital in Pune, India. Academic and private cardiology practices were then invited to participate voluntarily in PIQIP. PIQIP representatives conducted site visits at each of the interested practices to install standardized equipment and resources for data acquisition, including a scanner and Android-based datacollection software. PIQIP representatives also provided training to practice personnel for initial data capture. Sites from 4 major Indian cities (Ahmedabad, Mumbai, Pune, and Hyderabad) were represented in this registry [fig_ref] Figure 2: Map of India showing towns and cities where the Practice Innovation and... [/fig_ref].
Patients evaluated in outpatient departments (OPDs) in India usually receive an OPD card that lists basic demographic information, medical conditions, laboratory data, and related prescriptions. Unlike the PINNACLE registry in the United States, for which data collection is achieved through integration with practices' electronic medical records, [bib_ref] Cardiac performance measure compliance in outpatients: the American College of Cardiology and..., Chan [/bib_ref] data in PIQIP are collected primarily by trained personnel scanning each patient's OPD card and extracting the requisite information into the PIQIP Patient Care Tool and then uploading the information to a tablet computer that physicians can use to track their patients' progress on a longitudinal basis. [bib_ref] Thinking globally to transform cardiovascular care, O'gara [/bib_ref] The dataentry staff held bachelor degrees in pharmacy. A 3-tiered process was instituted to ensure data integrity. Tier 1 was frontline data entry, tier 2 involved quality assurance by sampling 25% of the charts, and tier 3 involved auditing 5% charts.
## Study population
For the purposes of this study, we included data on patients enrolled in PIQIP from January 1, 2011, to February 5, 2014. All patients from participating cardiology OPDs were eligible for the study, with 100% sampling from each participating OPD. The final study population included 68 196 unique patients from 10 cardiology OPDs, encompassing a total of 155 953 patient encounters. For this study, risk factors and disease states were defined based on physicians' documentation. Data on medication use in different CVD states represent use of medications by unique patients at any encounter.
## Statistical analyses
In the current study, we examined the prevalence of CAD, HF, and AF for the first 68 196 patients. We also provided basic demographic data, presence of related cardiovascular risk factors, and use of evidence-based medication for each disease condition. All analyses were conducted with SAS version 9.3 (SAS Institute).
The PINNACLE registry in the United States had a waiver of informed consent; although the PIQIP did not undergo institutional review board application in India, standard practices for data collection and patient data protection as in the United States, including handling of patient data in compliance with Health Insurance Portability and Accountability Act practices, were followed. The vendor conducting the data entry in the PIQIP was the same vendor used for the PINNACLE registry in the United States.
# Results
Of the 10 cardiology OPDs that participated in the study, 7 were private OPDs, 2 were run by trusts, and 1 had an academic affiliation with a medical school. Six of the 10 OPDs were based in major cities (Ahmedabad, Mumbai, Pune, and Hyderabad). Collectively, 115 cardiologists from these 10 OPDs submitted data to PIQIP.
There were a total of 68 196 unique study patients, encompassing a total of 155 953 patient encounters. The mean age of the study population was 50.6 years (SD 18.
## Cad risk factor profile in piqip
Hypertension was the most prevalent CAD risk factor, reported in 20 268 patients (29.7%), of which 14 400 (71.0%) were men. Diabetes mellitus was reported in 10 167 patients (14.9%), of which 7675 (75.0%) were men. Current tobacco use of any type was present in 5202 patients (7.6%), of which 4163 (80.0%) were men [fig_ref] Table 1: Prevalence of Cardiovascular Disease Risk Factors and Coronary Artery Disease, Heart Failure,... [/fig_ref].
## Profile of patients with cad and evidence-based medication use in cad patients
CAD was present in 10 106 patients (14.8%) who were predominantly male (8639 patients [85.5%]). The mean age of the CAD population was 58.3 years (SD 11.4 years), with a lower mean age of men represented (57.9 years [SD 11.3 years]) compared with women (60.9 years [SD 11.6 years]). Hypertension, diabetes mellitus, and current tobacco use of any type were present in 5124 (50.7%), 3327 (32.9%), and 342 (3.4%) patients with CAD, respectively.
Medication use among CAD patients is summarized in [fig_ref] Table 2: Medication Use for Coronary Artery Disease in the PINNACLE India Quality Improvement... [/fig_ref]. Aspirin was prescribed for 4908 CAD patients (48.6%), clopidogrel was prescribed for 3747 (37.1%), and prasugrel was prescribed for 696 (6.9%). Similarly, betablocker and RAAS (renin-angiotensin-aldosterone system) antagonist (including angiotensin-converting enzyme inhibitors and angiotensin receptor blockers) therapies were prescribed in 4762 (47.1%) and 4863 (48.1%) CAD patients, respectively. The use of lipid-lowering agents was documented in 5385 CAD patients (53.3%), with 5112 (50.6%) documented as being on a statin-based therapy and 273 (2.7%) documented as receiving non-statin-based lipid-lowering therapy.
## Profile of patients with hf and evidence-based medication use in hf patients
History of HF was present in 2752 patients (4.0%) who were predominantly male (2230 patients [81.0%]). The mean age of patients with HF was 56.9 years (SD 13.8 years).
Medication use among HF patients is summarized in [fig_ref] Table 3: Medication [/fig_ref]. Prescription for RAAS antagonist therapy was documented in 1704 patients (61.9%), beta-blocker use was documented in 1599 (58.1%), and diuretic use was documented in 1384 (50.3%).
## Profile of patients with af and af medication use
History of AF was documented in 348 patients (0.5%). The prevalence of AF was 0.3% (163 patients) in men and 0.9% (178 patients) in women. The mean age of the study population with AF was 56. . All patients with AF who were also prescribed clopidogrel had concomitant CAD. Medication use among AF patients is summarized in [fig_ref] Table 4: Medication [/fig_ref].
# Discussion
PIQIP is the first outpatient CVD-care data-collection and reporting program implemented in India. In a country with a disproportionate provider/patient ratio 10 and low levels of government funding for quality improvement, physician-driven initiatives for practice-based learning and improvement could be considered an unrealistic expectation. 4 PIQIP demonstrates the feasibility of data acquisition in this challenging environment.
In these preliminary analyses from the PIQIP registry, we examined clinical profiles of the first 68 196 CVD outpatients, encompassing 155 953 patient encounters from 10 cardiology OPDs in India. There were several important findings. Although there seemed to be a significant difference by sex among CVD encounters, with women composing only 7% and 3% of patients with CAD and HF, respectively, more data are required before a firm conclusion on gaps in access to specialty care by sex can be made. Other significant observations from our pilot study include younger mean age of the represented populations with CVD and relatively lower prescription of evidence-based medications for CAD, HF, and AF. Only a large-scale nationwide study in India with adoption of several methods to overcome the barriers that we have encountered with our data collection and use of quality-control measures to ensure data integrity can provide better evidence about cardiovascular demographics and evidence-based medication use in India. It is possible that local documentation practices could explain the lower medication prescription and prevalence of tobacco use seen in our study, especially because robust medical record keeping in India is not common. A lower prevalence of tobacco use in our cohort compared with previously published data 11 could also reflect other differences such as a higher rate of smoking cessation in patients with established CVD and variation based on the specific study population (ie, patients seeking care in cardiology OPDs in India in our study compared with the general Indian population). Nevertheless, our study shows the feasibility of initiating such a novel quality-improvement program in a resource-limited setting such as India.
Unique challenges were related to the implementation of PIQIP. A majority of the OPDs participating in the registry did not use electronic medical records. Although health information management sections of hospitals store paper charts for inpatients following hospitalization for a specific duration before records are destroyed, medical record keeping for outpatients in India is mostly nonexistent. Patients who were evaluated in the OPDs in hospitals or offsite clinics usually received an OPD card that they were responsible for and brought to their OPD visits. Each outpatient visit usually generated a new OPD card, regardless of whether the index visit was an initial or follow-up appointment. Consequently, linking data from one visit to another was difficult, given the lack of unique identification of patients from visit to visit. Given their clinical responsibilities, including high patient volume, physicians were reluctant to change their workflow and to use the Web-based tool, adding further complexity to data acquisition; however, physicians were ready to scan prescription or OPD cards and send them to the ACC site in Pune for data entry. In addition, ancillary staff responsible for scanning prescriptions or OPD cards required repeated training to operate the software appropriately, and that put increased demand on resource availability and funding. Finally, prescriptions or OPD cards were hand-written by physicians and were not always legible, making data extraction difficult. With wider implementation of electronic medical records in India in the future, we expect some of these barriers to be less problematic. Moreover, careful resource allocation to improve documentation metrics without adversely affecting clinical work flow will be an important step to improve the quality of cardiovascular care in India. Several strategies are being considered for the future course of the PIQIP registry: progression to a model that enables OPDs to become self-sufficient in data collection and reporting and, consequently, less dependent on ACC staff; enhanced data capture including socioeconomic variables, medication contraindication, and laboratory values; and expansion of the program to other sites across India. If feasible, to improve data integrity, we will consider additional measures in the future, such as validation for the presence of risk factors such as hypertension and diabetes by examining measured blood pressure, blood glucose (or hemoglobin A1c), and medication use for such conditions in a random sample of patients from each site. Similarly, presence of cardiac diagnoses can be validated by examining documentation of coronary angiograms or noninvasive tests for CAD, echocardiogram for HF, and electrocardiogram for atrial fibrillation in a randomly selected sample. These efforts at data collection and validation of cardiovascular risk factors and disease conditions will be coordinated closely with local cardiologists.
This study has several important limitations. First, patients were followed longitudinally only if they returned to a data-collecting OPD; in other words, longitudinal data could be collected only for patients who returned to a site that they had visited before. Second, the study relied on clinicians for data collection; therefore, the success of the program was contingent on clinicians' diligence regarding medical record documentation, which may not be complete. Contraindication to an indicated medication, for example, was not captured by the PIQIP registry but could account for the lower rate of medication prescriptions. In addition, it is possible that there was ascertainment bias related to capture of risk factors (ie, hypertension, diabetes mellitus, and current tobacco use) in this registry. Furthermore, CHADS2 score was used for nonvalvular AF, and data on etiology of AF were not available. The World Bank data from India demonstrates that the average doctor-patient interaction lasts for 3.8 minutes, in which the doctor asks 3.2 questions and performs slightly >1 physical examination procedure. [bib_ref] The quality of medical advice in low-income countries, Das [/bib_ref] Traditionally, physicians in India document only the diagnosis followed by a list of prescription medications on OPD cards, unlike US office visit documentation, which comprises a detailed history and physical examination, medications, allergies, physical examination, laboratory and diagnostic data, and assessment and plan information. Although data-collection forms were designed to encourage improved documentation and capture of all key encounter variables, physicians could not be compelled to document missing diagnoses, laboratory values, or medications. Furthermore, despite the data being extracted by trained individuals, data were not reviewed later by an independent physician or research nurse. Consequently, the lower reported use of evidence-based medications in this registry could reflect the local documentation practices in India. Third, given their participation in a quality-improvement registry, the participating practices could represent highly motivated sites, and thus these data may not be generalizable to the larger Indian population with CVD. Finally, it is possible that the cost associated with medications could influence the lower prescription rate in India. Despite these limitations, the PIQIP registry collected real-world data. This approach was advantageous for the swiftness of setup and for scalability, sustainability, and cost.
# Conclusion
In this first report from the PIQIP registry, we showed the feasibility of studying outpatient quality of cardiovascular care in India. In addition, this report describes several challenges unique to outpatient quality measurement in India and outlines future strategies to overcome them.
[fig] Figure 1: Infographic demonstrating, in a chronological fashion, the data-collection steps that were implemented by the American College of Cardiology for the PINNACLE India Quality Improvement Program. ACC indicates American College of Cardiology; AF, atrial fibrillation; CAD, coronary artery disease; HF, heart failure; OPD, outpatient department; PINNACLE, Practice Innovation and Clinical Excellence; RAAS, reninangiotensin-aldosterone system. [/fig]
[fig] 2: years). A majority of the represented population was aged <65 years (53 497 patients [78.5%]) and was male (47 697 patients [70.0%]). The mean blood pressure was 126/ 78 mm Hg (SD 20/11 mm Hg), and the mean body mass index was 25.8 (SD 4.8) in kg/m 2 . [/fig]
[table] 1: years (SD 14.9 years). A majority of these patients were aged <65 years (239 [68.7%]) compared with patients aged 65 to 74 years (74 [21.3%]) or ≥75 years (35 [10.0%]). Hypertension was present in 39.6% (138 patients), followed by diabetes mellitus in 16.4% (57 patients) and HF in 10% (34 patients). CHADS2 score was 0 for 179 patients (51.4%), 1 for 105 patients (30.2%), 2 for 55 patients (15.8%), 3 for 8 patients (2.3%), and 4 for 1 patient. Beta-blocker and calcium-channel antagonist use was documented in 147 (42.2%) and 26 (7.5%) patients, respectively. Similarly, use of systemic anticoagulation with oral vitamin K antagonist (warfarin) or one of the novel oral anticoagulants (dabigatran) was documented in 129 patients (37.0%), with a vast majority (112 patients, 87.0%) prescribed warfarin therapy. Antiplatelet agents concomitantly used in patients with atrial fibrillation included aspirin (77 patients [22.1%]) and clopidogrel (40 patients [/table]
[table] Table 3: Medication [/table]
[table] Table 4: Medication [/table]
|
Missed Abortion Presented with Worsening Hyperemesis Gravidarum
# Introduction
Hyperemesis gravidarum (HG) accounts for the highest number of in-patient hospitalizations during the first half of pregnancy. The pathogenesis of HG remains unclear so that the treatment is mainly supportive. HG can lead to several serious maternal complications, including severe dehydration, nutritional deficiency, and multiple electrolytes imbalances, which can result in cardiac arrhythmia and death. However, HG-related fetal complications remain contradictory. Some studies suggested that HG causes abnormal placentation leading to pre-eclampsia and small for gestational age (GA) of the fetus. To date, there is no conclusive evidence support whether HG directly increases the risk of abortion. In fact, patients with abortion are likely to have resolving HG due to a drop in gestational hormones. We report a unique case of worsening HG as a presentation of a missed abortion.
## Case presentation
A 32-year-old nulliparous woman with GA of 14 weeks came to the hospital due to worsening nausea and vomiting for two days. She had had nausea and vomiting for approximately a month, without apparent causes, which was diagnosed as HG and received symptomatic treatment. She had no other medical problems nor any other complaints. On initial evaluation, she had tachycardia and significant dehydration without fever. Her blood tests were abnormal for hyponatremia with sodium of 131 millimoles per liter (mmol/L), hypokalemia with potassium of 2.8 mmol/L, and hypophosphatemia with phosphorus of less than 0.3 mmol/L. She also had elevated creatinine at 1.1 milligrams per deciliter (mg/dl) (her baseline creatinine was 0.52 mg/dl) and metabolic acidosis with an anion gap of 32 due to starvation ketoacidosis. Her electrocardiogram (ECG) showed sinus tachycardia with a heart rate of 135 beats per minute and marked ST abnormalities .
## Figure 1: electrocardiogram showed sinus tachycardia with a heart rate of 135 beats per minute and marked st abnormalities
Ultrasound pelvis illustrated an abnormal appearing fetus without a cardiac activity, estimated GA of 12 weeks, and four days, compared to the previous ultrasound study a month prior, showed GA of 10 weeks and one day, indicating an interval missed abortion.
She was admitted to a medical intensive care unit, received aggressive electrolyte replacement and intravenous crystalloid bolus, and underwent dilatation and evacuation to remove the conceptus. Her postoperative course was uneventful. Her hospital course was complicated by an adjustment disorder with depressed mood, which was managed by supportive psychotherapy. Her conditions were improved, and she was discharged home on day 5.
# Discussion
HG is one of the most devastating forms of nausea and vomiting during pregnancy, leading to serious maternal morbidity and death. However, HG related to adverse fetal outcomes remains controversial. HG may result in low birth weight and pre-eclampsia. We presented a case of severe HG, leading to several critical electrolytes abnormalities, significant dehydration, and acute kidney injury. In our patient, prompt recognition and treatment of HGrelated complications leads to favorable outcomes and reduces hospital length of stay. HGrelated maternal complications requires urgent management to prevent catastrophic outcomes, such as cardiac arrhythmia, renal failure, nutritional deficiency, and death.
Patients suffering from HG presented with frequent vomiting and unable to tolerate oral intake, which will result in dehydration, metabolic alkalosis, electrolyte abnormalities, and nutritional deficits. These might result in life-threatening cardiac arrhythmias, seizures, acute kidney injury, osmotic demyelination syndrome, and Wernicke encephalopathy. Severe vomiting can also lead to fatal esophageal rupture and pneumomediastinum.
The differential diagnosis of HG includes gastrointestinal tract disorders, hepatobiliary diseases, pyelonephritis, and endocrine abnormalities. HG is a diagnosis of exclusion without any valid diagnostic markers. The treatment of HG is mainly supportive: the replacement of fluid and electrolytes and administration of antiemetic medications.
Our patient was also found to have missed abortion, which may point out that severe HG can lead to fetal death and abortion. However, more evidence is needed to support our hypothesis.
# Conclusions
HG is a common and potentially lethal condition during pregnancy. Clinicians should be aware of HG-related complications. Our case report presentation suggested that severe HG may also result in abortion.
# Additional information disclosures
Human subjects: Consent was obtained by all participants in this study.
## Conflicts of interest:
In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work. |
A randomized placebo-controlled trial of convalescent plasma for adults hospitalized with COVID-19 pneumonia
Passive immunotherapy with convalescent plasma may be the only available agent during the early phases of a pandemic. Here, we report safety and efficacy of high-titer convalescent plasma for COVID-19 pneumonia. Double-blinded randomized multicenter placebo-controlled trial of adult patients hospitalized with COVID-19 pneumonia. The intervention was COVID-19 convalescent plasma and placebo was saline allocated 2:1. The primary outcome was clinical status 14 days after the intervention evaluated on a clinical ordinal scale. The trial was registered at ClinicalTrials.Gov, NCT04345289, 14/04/2020. The CCAP-2 trial was terminated prematurely due to futility. Of 147 patients randomized, we included 144 patients in the modified intention-to-treat population. The ordinal clinical status 14 days post-intervention was comparable between treatment groups (odds ratio (OR) 1.41, 95% confidence interval (CI) 0.72-2.09). Results were consistent when evaluating clinical progression on an individual level 14 days after intervention (OR 1.09; 95% CI 0.46-1.73). No significant differences in length of hospital stay, admission to ICU, frequency of severe adverse events or all-cause mortality during follow-up were found between the intervention and the placebo group. Infusion of convalescent plasma did not influence clinical progression, survival or length of hospitalization in patients with COVID-19 pneumonia.
## Scientific reports
| (2022) 12:16385 | https://doi.org/10.1038/s41598-022-19629-z www.nature.com/scientificreports/ has been used in the past century for respiratory infections caused by SARS-CoV-1, middle eastern respiratory syndrome (MERS), and influenza virus with uncertain benefit [bib_ref] Meta-analysis: Convalescent blood products for Spanish influenza pneumonia: A future H5N1 treatment?, Luke [/bib_ref] [bib_ref] Convalescent plasma treatment reduced mortality in patients with severe pandemic influenza A..., Hung [/bib_ref] [bib_ref] The effectiveness of convalescent plasma and hyperimmune immunoglobulin for the treatment of..., Mair-Jenkins [/bib_ref]. Several randomized controlled trials of COVID-19 convalescent plasma were initiated during the first year of the pandemic, including this trial [bib_ref] Association of convalescent plasma treatment with clinical outcomes in patients with COVID-19:..., Janiaud [/bib_ref]. The largest trial of convalescent plasma, the Randomized Evaluation of COVID-19 Therapy (RECOVERY) Trial, enrolled more than 11,000 participants. In RECOVERY, high-titer convalescent plasma did not improve survival or other clinical outcomes. These results were published as a preprint on March 10, 2021 [bib_ref] Convalescent plasma in patients admitted to hospital with COVID-19 (RECOVERY): A randomised..., Recovery Collaborative Group [/bib_ref]. Based on these results, our Data and Safety Monitoring Board (DSMB) recommended stopping enrolment to our trial, the CREDID COVID-19 Adaptive Platform (CCAP)-2 trial, on due to a high likelihood of futility.
The objective of our trial was to evaluate the efficacy and safety of convalescent plasma compared with placebo for hospitalized patients with COVID-19 pneumonia. Here we report the modified intention-to-treat (mITT) analysis of the primary and secondary outcomes of the individuals who received the allocated trial intervention in CCAP-2.
# Methods
Study design. The CREDID COIVD-19 Adaptive Platform (CCAP)2 trial was a randomized doubleblinded, placebo-controlled, multicenter trial conducted at six clinical sites in Denmark. Eligible participants were randomly assigned in a 2:1 ratio to receive convalescent plasma or placebo. The trial protocol was approved by the Ethical Committee of the Capital Region of Denmark (record no. and is available in the Appendix. Written informed consent was obtained from all participants, and the trial was conducted in accordance with the principles stated in the Declaration of Helsinki. The authors take full responsibility for the design and conduct of the trial and vouch for the accuracy and completeness of the data, the analysis of the data, and the adherence of the trial to the protocol. The trial was registered at ClinicalTrials.Gov, NCT04345289, 14/04/2020.
## Participants.
At each participating site, hospitalized adults (≥ 18 years) were eligible for enrolment if they were positive for SARS-CoV-2 by a reverse-transcriptase-polymerase-chain-reaction (RT-PCR) of a respiratory tract sample, had evidence of pneumonia as determined by radiologically confirmed pulmonary infiltrate and/ or peripheral oxygen saturation (SaO2) below 94% while breathing ambient air, and a negative pregnancy test for women of childbearing potential.
Exclusion criteria of the study were I) Pregnancy or breastfeeding, II) Participation in other drug clinical trials, III) Any serious medical condition or abnormality of clinical laboratory test, that in the investigator's judgement precluded the patients safe participation or completion of the study, IV) Progression to death was imminent and inevitable within the next 24 h, irrespectively of the provision of the treatment.
## Randomization and masking. eligible patients underwent treatment allocation and concealment
through the randomization program Research Electronic Data Capture (REDCap) and were assigned in a 2:1 ratio to receive either COVID-19 convalescent plasma (600 mL) or placebo (600 mL of 0.9% saline solution) in addition to standard of care treatment [bib_ref] The REDCap consortium: Building an international community of software platform partners, Harris [/bib_ref] [bib_ref] Research electronic data capture (REDCap): A metadata-driven methodology and workflow process for..., Harris [/bib_ref]. Opaque covers were used to conceal the intravenous lines and infusion bags (Supplemental [fig_ref] Figure 1: Enrollmentcharacteristics of patients included in the mITT analysis are shown in Table... [/fig_ref]. The participants, the clinical team, the data collectors, and the outcome adjudicators were all blinded to the treatment assignments. Patients could receive remdesivir, glucocorticoids and anticoagulants according to the standard of care at the providing health care institution.
Trial procedures. Plasma donors were males between 18 and 65 years of age with a previously confirmed SARS-CoV-2 infection (presence of SARS-CoV-2 nucleic acid by PCR or anti-SARS-CoV-2 immunoglobulin), who had been asymptomatic for at least 4 weeks at time of plasmapheresis, and met general criteria for plasma donor eligibility. Male donors were used in order to reduce the risk of Transfusion-related acute lung injury (TRALI) due to anti-human leukocyte antigen antibodies that are more frequent in female donors. Plasma donors were either registered blood donors or previous patients diagnosed with COVID-19. All plasma donors provided informed consent prior to plasmapheresis as per local blood donor protocol. Plasma was obtained by plasmapheresis at the blood establishment associated with the public hospital services in the five Danish Regions according to standard procedures and national guidelines 8 . At each apheresis, 600 mL of plasma was collected and split into two units of 300 mL. Convalescent plasma was required to have a minimum SARS-CoV-2 immunoglobulin (Ig) G antibody titer of > 3.0 (EUROIMMUN AG, Lübeck, Germany). Convalescent plasma for each patient was predominantly from a single donor. Both units of 300 mL were transfused within 24 h.
Outcomes. The primary outcome of the trial was changed from a comparison of survival on day 28 to clinical status on day 14. The amendment was proposed on January 15, 2021, after discussions between the DSMB and sponsor due to low study accrual and a nationwide decline in new cases. The trial was paused until the amendment was approved on January 23, 2021. The original primary outcome became the key secondary end point. The revised primary outcome of clinical status 14 days after intervention was defined by one of seven mutually exclusive ordinal categories on an adapted version of the World Health Organization (WHO) clinical scale: (1) death; (2) hospitalized, in intensive care requiring Extracorporeal Membrane Oxygenation (ECMO) or mechanical ventilation; (3) hospitalized, on non-invasive ventilation or high-flow oxygen device; (4) hospitalized, requiring supplemental oxygen; (5) hospitalized, not requiring supplemental oxygen; (6) not hospitalized, limitations on activities and/or requiring home oxygen; and (7) not hospitalized, no limitations on activities [bib_ref] A minimal common outcome measure set for COVID-19 clinical research, Marshall [/bib_ref]. On day 14, the primary outcome was determined as the worst of the seven categories which the participant fulfilled.
Secondary outcomes were: Frequency of infusion-related adverse events, frequency of severe-adverse-events, all-cause mortality or need of invasive mechanical ventilation up to day 28, ventilator-free days to day 28, organ www.nature.com/scientificreports/ failure-free days to day 28, duration of intensive care unit (ICU) stay, mortality at days 7, 14, 21, 28, and 90, length of hospital stay and duration of supplemental oxygen. After the trial intervention, patients were followed in person during the hospital admission and by telephone after hospital discharge. Details regarding data collection, patient follow-up, randomization, the data blinding and masking process, and plasma donation, collection, processing, and storage are provided in the Supplementary.
# Statistical analysis.
With the original primary outcome, the trial was designed to enroll 1,100 participants.
The change in primary outcome permitted a more realistic sample size with use of a widely accepted clinical outcome used for COVID-19 trials. The amendment was proposed when 111 of 1100 participants according to the original sample size calculation had been enrolled. In the revised version, the trial was designed to enroll 530 patients (353 in the plasma group and 177 in the placebo group) assuming a specific severity distribution at day 14, including an expected mortality rate within 14 days of 6.5% in the intervention group and 10% in the placebo group (Supplemental [fig_ref] Table 1: Baseline characteristics of patients included in the modified intention to treat analysis [/fig_ref]. These assumptions were based on data from studies conducted in the early phase of the pandemic [bib_ref] Remdesivir for the treatment of Covid-19: Final report, Beigel [/bib_ref] [bib_ref] Effect of hydroxychloroquine on clinical status at 14 days in hospitalized patients..., Self [/bib_ref]. We calculated that this sample size would provide 80% power to detect a proportional odds ratio of 1.6 for plasma as compared with placebo on the clinical ordinal scale at the 0.05 (two-sided) level of significance. The analysis was performed using R version 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria). Details regarding the overall statistical analysis, interim analysis, and unblinding criteria are provided in the Supplementary Material.
# Results
Patients. Between June 13, 2020 and March 16, 2021, a total of 289 patients were assessed for eligibility, and 147 patients were enrolled. Three patients withdrew informed consent before receiving their treatment intervention. Consequently, 98 patients received convalescent plasma and 46 participants received placebo [fig_ref] Figure 1: Enrollmentcharacteristics of patients included in the mITT analysis are shown in Table... [/fig_ref] and were included in the mITT analysis. Two participants, who were discharged on day 2 and 3, respectively, refused contact during follow up but were alive on day 90. Their ordinal scale status was conservatively set to 6 on day 14. The majority (92%) of the patients received supplemental oxygen, of whom 5% required mechanical ventilation, 33% high-flow oxygen or non-invasive ventilation, and 54% low-flow oxygen. Most patients were treated with remdesivir (87%), dexamethasone (90%), and anticoagulation (55%) at the time of trial entry. www.nature.com/scientificreports/ At baseline, patients receiving convalescent plasma more often required high-flow oxygen or non-invasive ventilation than patients who received placebo [fig_ref] Table 1: Baseline characteristics of patients included in the modified intention to treat analysis [/fig_ref] and [fig_ref] Figure 2: Clinical progression scale at baseline [/fig_ref]. Patients receiving convalescent plasma and placebo were comparable on all other baseline variables.
Intervention and placebo administration. The median EUROIMMUN SARS-CoV-2 IgG titer of the infused convalescent plasma was 5.2 (IQR 3.6 to 6.6). All participants in the placebo group received the full saline infusion whereas 93 of 96 in the intervention group received the full amount of convalescent plasma; one discontinued due to hypotension and two because of a skin rash.
## Primary outcome. the distribution of clinical outcomes on day 14 according to the clinical ordinal scale
was comparable between the convalescent plasma group and the placebo group, odds ratio (OR) 1.41; 95% confidence interval [CI], 0.72 to 2.09; P = 0.321) for worse outcome [fig_ref] Table 2: Worst scores for Covid-19 severity at 14 days according to a 7-category... [/fig_ref] and [fig_ref] Figure 2: Clinical progression scale at baseline [/fig_ref]. In a post hoc analysis, individual clinical progression by day 14 did not differ according to treatment group (OR 1.09; 95% CI, 0.46-1.73) for convalescent plasma vs. placebo [fig_ref] Figure 3: Clinical progression scale on individual level by 14 days [/fig_ref].
Secondary outcomes. The 90-day mortality was 16% (15 of 96 patients) in the convalescent plasma group and 9% (4 of 46) in the placebo group . No significant between-group differences in survival was seen at day 7, 14, 28, or 90. The median time from enrollment to death was 11 days 5-14 in the convalescent plasma group and 9 days 8-12 in the placebo group. Thirteen of 16 deaths in the convalescent plasma group and four out of four deaths in the placebo group occurred during hospitalization for COVID-19. ICU admission and mechanical ventilation was required by 14% and 7%, respectively, in the convalescent plasma group and 10% and 5%, respectively, in the placebo group. Duration of ICU admission was seven days (IQR 4-18) for patients in the convalescent plasma group compared to 14 days (IQR 8-20) for patients in the placebo group. No differences were noted for ventilator-free days, organ-failure-free days or oxygen supply days between groups. The median time from enrollment to hospital discharge was 6 days (IQR 3-13) in the convalescent plasma group and four days (IQR 2-8) in the placebo group. www.nature.com/scientificreports/ Safety results. Infusion-related adverse events (hypotension (n = 1), rash (n = 2) and respiratory distress (n = 1)) were more common in the convalescent plasma group (4.2%; 4 of 96 patients) than in the placebo group (hypotension (n = 1) (2.2%; 1 of 46 patients). Although statistically insignificant, a higher frequency of severe adverse events was reported for the convalescent plasma group compared to the placebo group (26% vs 17%). Reported adverse and severe adverse events during follow-up are provided in [fig_ref] Table 2: Worst scores for Covid-19 severity at 14 days according to a 7-category... [/fig_ref].
# Discussion
The CCAP-2 trial was stopped after a third of participants had been recruited based on a recommendation from the DSMB due to futility. In this prematurely terminated trial, we found that hospitalized patients with COVID-19 pneumonia who received convalescent plasma did not have better clinical outcomes at day 14 than those who received placebo. Similarly, the rates of death through 90 days as well as other secondary outcomes were comparable in both treatment arms. The safety profile was comparable with convalescent plasma and placebo. Numerically, there was a higher proportion of individuals with more severe ordinal outcomes at day 14 and more severe adverse events and deaths at all time points in the convalescent plasma group in our study. The differences were not statistically significantly different at any point. By chance, participants in the convalescent plasma group had significantly more severe disease at inclusion by the ordinal scale. It is likely, that this explains . Secondary outcomes. SAE Severe adverse events, IQR interquartile range, ICU Intensive care unit, CI confidence interval. *Patients on mechanical ventilation at randomization not included. **Patients admitted to the ICU at randomization not included. ***Patients not requiring oxygen supply not included.
## Convalescent plasma (n = 98) placebo group (n = 46)
Odds ratio (95% CI) or median difference (95% CI) Ventilator-free days through 28 days, median (IQR) 28 (28-28) 28 (28-28) 0 (0-0)
[formula] Deaths [/formula]
Organ failure through 28 days (%) 9/98 (9) 3/46 (7) www.nature.com/scientificreports/ the differences in outcomes between groups because an analysis of individual outcomes showed an OR of 1.05 (95% CI, 0.41-1.69) for clinical progression by day 14 (convalescent plasma versus placebo). The uncertainty of the 28-day mortality estimate in this study encompasses the results reported by the RECOVERY trial and an accompanying metaanalysis of 11 trials including an additional 1,973 individuals to the 11,258 participants of RECOVERY, showing no beneficial effects of treatment with convalescent plasma in relation to mortality at day 28 5 . The RECOVERY trial did not report clinical progression by day 14.
Convalescent plasma is administered to increase levels of neutralizing antibodies in the recipient before the recipient has mounted a humoral response. Trial data suggests that administration of plasma within 72 h of symptom onset may prevent progression to severe respiratory disease for outpatients with mild disease, while administration to high-risk outpatients within 7 days of symptom onset in another trial did not prevent disease progression [bib_ref] Early high-titer plasma therapy to prevent severe covid-19 in older adults, Libster [/bib_ref] [bib_ref] Early convalescent plasma for high-risk outpatients with Covid-19, Korley [/bib_ref]. A randomized trial from Spain (n = 350) reported a benefit of convalescent plasma given within 7 days of symptom onset for a composite endpoint including respiratory status and death at day 28 [bib_ref] A multicenter randomized open-label clinical trial for convalescent plasma in patients hospitalized..., Avendaño-Solá [/bib_ref]. Early administration of high-titer plasma was associated with a lower risk of death in an observational trial [bib_ref] Convalescent plasma antibody levels and the risk of death from Covid-19, Joyner [/bib_ref]. In our trial, participants were enrolled into the trial within a median of 2 days (IQR 1.8-4.0) of admission to hospital and regardless of length of symptoms. The median duration of symptoms was 11 days (IQR 8-11) at randomization in our trial and, thus, comparable to most other trials that did not show a benefit of convalescent plasma [bib_ref] Convalescent plasma in patients admitted to hospital with COVID-19 (RECOVERY): A randomised..., Recovery Collaborative Group [/bib_ref] [bib_ref] Convalescent plasma in the management of moderate covid-19 in adults in India:..., Agarwal [/bib_ref] [bib_ref] A randomized trial of convalescent plasma in Covid-19 severe pneumonia, Simonovich [/bib_ref]. Therefore, convalescent plasma may still play a role if administered early and before the development of native antibodies. To further investigate a strategy of early administration in a controlled trial, participants would need to be recruited prior to hospital admission and preferably also have their antibody levels determined at baseline.
Any effect of convalescent plasma is most likely correlated with the titers of neutralizing antibodies, and failure to demonstrate a beneficial effect of convalescent plasma in prior studies could potentially be a consequence of use of plasma with low titers of neutralizing antibodies. In this study, we used convalescent plasma with a EUROIMMUN IgG titer > 3.0, which correlated closely (R 2 = 0.91) to an IgG titer of 15 measured by the Ortho Vitros assay [bib_ref] Comparison of 16 serological SARS-CoV-2 immunoassays in 16 clinical laboratories, Harritshøj [/bib_ref]. Hence the convalescent plasma used in this study qualified as high-titer plasma in accordance with the US FDA Emergency Use Authorization for COVID-19 convalescent plasma of August, 2020, which required an anti-SARS-CoV-2 IgG titer of 12 measured by the Ortho Vitros assay. The median EUROIMMUN IgG titer was in fact > 5 in this trial and, as such, it is unlikely that the null result observed in this study was related to insufficient titers of neutralizing antibodies.
Study amendments. The trial design was amended on January 15, 2020. The amendment included changing the primary outcome from a comparison of survival on day 28 to clinical status on day 14. This was done for several reasons. Accrual to the study had been low initially due to few hospitalizations between June and October of 2020, hospitals were overwhelmed in November and December of 2020 due to a steep increase in hospitalizations leading to transfer of study resources at several study sites (in fact, only three of six sites enrolled a significant number of participants), there was a temporary shortage of convalescent plasma in the Capital Region of Denmark (the region had enrolled two thirds of participants) in January 2021 due to limited resources, and a waning epidemic after January 2021with fewer hospitalizations for COVID-19. The advice from our Data Safety and Monitoring Board was that enrollment of 1000 participants wasn't feasible. Additionally, an increasing number of COVID-19 trials applied the revised WHO clinical ordinal scale as the primary outcome to evaluate efficacy of new treatments. Thus, this specific outcome has become widely accepted as an appropriate outcome measure in COVID-19 trials. By adopting this outcome to our study, we allowed for better comparison with similar studies on the subject. In addition, the revised outcome permitted a feasible sample size.
# Strengths and limitations.
Strength of our trial include the double-blinded and placebo-controlled multicenter design, few dropouts and near complete follow up. Moreover, only high-titer convalescent plasma was used in the study. Limitations included a high rate of ineligibility that may reduce the external validity as well as insufficient sample size and statistical power owing to the premature termination of the trial.
# Conclusion
In conclusion, the infusion of convalescent plasma did not influence clinical progression, survival, or length of hospitalization in patients with COVID-19 pneumonia when administered during hospital admission.
## Data availability
Data will not be made available for download due to restrictions regarding privacy under the EU GDPR. Access to data may be granted on specific requestion to [email protected] and only in agreement with the rules and regulations of the Danish Data Protection Agency and the National Committee on Health Research Ethics.
[fig] Figure 1: Enrollmentcharacteristics of patients included in the mITT analysis are shown in Table 1. The median age of the patient population was 65 years (interquartile range [IQR] 55-75); 72% of the patients were men, and 77% had a coexisting comorbidity at trial entry. The median time from the onset of COVID-19 symptoms to enrollment was 11 days (IQR 8-13). At baseline, patients had a median respiratory rate of 20 per min (IQR 18-22) and a median oxygen saturation of 94% (IQR 92-96). Median values of lymphocytes, D-dimer and C-reactive protein were 0.8 × 10 9 (IQR 0.7-1.2) cells/L, 0.8 (IQR 0.5-1.4) Units/L and 64.5 (IQR 36.2-115.5) mg/L, respectively. [/fig]
[fig] Figure 2: Clinical progression scale at baseline (A) and by 14 days (B). [/fig]
[fig] Figure 3: Clinical progression scale on individual level by 14 days. [/fig]
[table] Table 1: Baseline characteristics of patients included in the modified intention to treat analysis. IQR interquartile range, NIV non-invasive ventilation, HF high-flow, P plasma, B blood. [/table]
[table] Table 2: Worst scores for Covid-19 severity at 14 days according to a 7-category ordinal scale. [/table]
|
Re-evaluation of retested nucleic acid-positive cases in recovered COVID-19 patients: Report from a designated transfer hospital in Chongqing, China
## A b s t r a c t
Since the outbreak of coronavirus disease 2019 in Wuhan, Hubei Province, China [1], a large number of confirmed cases met the discharge criteria (one of which is two consecutive negative nucleic acid tests with an interval of at least 24 h). Previous studies have paid more attention to the epidemic situation of COVID-19 and patient diagnosis and treatment. Close attention also should be paid to the discharged patients. Surprisingly, a previous follow-up reported that some patients' nucleic acid retest results were positive again after discharge [bib_ref] Positive RT-PCR test results in patients recovered from COVID-19, Lan [/bib_ref]. Factors impacting these follow-up test results should be further investigated. Since the first confirmed case was diagnosed in our hospital (Chongqing Emergency Medical Center, the designated transfer hospital) on February 4th, we confirmed a total of 17 cases. All patients infected with the novel coronavirus were transferred to a designated hospital in Southwest China's Chongqing by ambulance with an inbuilt negative-pressure chamber. In the follow-up examination of these patients, RT-PCR tests were conducted again 3 days after discharged from the designated hospital. Four patients showed recurrence of positive results after a few days of discharge. Thus, we examined these cases herein, aiming to provide information for policy formulation and modification of discharge plans.
© 2020 The Author(s). Published by Elsevier Ltd on behalf of King Saud Bin Abdulaziz University for Health Sciences. This is an open access article under the CC BY-NC-ND license (http://creativecommons. org/licenses/by-nc-nd/4.0/).
# Methods
Data were collected from the medical records of these patients with novel coronavirus pneumonia confirmed by Chongqing Emergency Medical Center. Their families, the Center for Disease Control and Prevention of Yuzhong District, Chongqing, and the designated hospital were contacted directly. We further analyzed the epidemiological history, clinical symptoms and multiple RT-PCR test results of virus nucleic acids after diagnosis and recovery.
All of the cases reported were confirmed in our hospital by the trial version 5 of the new coronavirus pneumonia diagnosis and treatment program. The patients were then transferred to the designated hospital to undergo treatment until they met the discharge criteria. The discharge decisions were made by a panel of experts on COVID-19. The patients were quarantined sequentially, and the nucleic acid test was conducted 3 days after discharge. All patients were followed up by telephone.
This study was approved by the medical ethics committee of the Chongqing Emergency Medical Center (No. 2020-06). Written informed consent was waived for emergency medical rescue of public health emergencies in accordance with the relevant provisions of national medical ethics.
[formula] + − − Discharge 33.15/Negative − − − − ND Nasopharyngeal swab (Ct of N/1ab) NA − − 32.77/33.83 − − − − − Anal swab (Ct of N/1ab) NA − NA 35.77/Negative − − − − ND+ − − Discharge 36/Negative − − − − − Nasopharyngeal swab (Ct of N/1ab) NA − − ND − − − ND ND Anal swab (Ct of N/1ab) NA NA NA 36.76/35.42 − NA NA ND ND [/formula]
# Results
Two patients showed positive nasopharyngeal swabs three days after discharge. The remaining patients showed positive anal swab results three days after discharge. However, the symptoms and CT manifestations of the disease were not exacerbated for the latter patient. Therefore, the former three patients returned to the designated hospital to be quarantined again. Two patients were discharged again from the hospital on March 2, 2020, and the nucleic acid was still negative after discharge. The other patient (case 4) is still under medical observation. Case 3 was quarantined in our hospital due to positive anal swab results.
## Case 1
A 29-year-old male, He is working in the service industry. The specific exposure history is unknown. This patient may represent the first case of a family cluster with the COVID-19. He lives with his parents, who were also confirmed to be infected with the novel coronavirus. He had symptoms of fever and cough at onset on January 27, and presented to our hospital 3 days later. Due to a positive nucleic acid test, he was confirmed as having novel coronavirus disease. After 14 days of treatment, he met the discharge criteria on February 15. The nasopharyngeal swab was positive 3 days after discharge, but it was subsequently negative for four times until March 5th, 2020.
## Case 2
A 49-year-old female (the mother of case 1) was exposed to her infected son. She had symptoms of cough and the confirmation test was made by a combination of exposure history and a positive nucleic acid test. She also received antivirus treatment for 14 days, until discharge on February 16th. Three days after discharge, the nasopharyngeal swab test result was positive. Four subsequent results were negative until March 5th, 2020.
## Case 3
A 12-year-old female was exposed to her infected mother. Thus, she was asked to visit our hospital to undergo oropharyngeal swabs for the RT-PCR test and to undergo a chest computed tomography (CT) examination. Mild COVID-19 was diagnosed due to positive RT-PCR tests results. Antivirus treatment with arbidol, interferon and traditional Chinese medicine was conducted in the designated hospital over the next 15 days. Then, she was met the discharge criteria (2 negative RT-PCR test results at least 1 day apart). In addition, she was asymptomatic, and the CT scan was negative. The nucleic acid test of anal swab was positive three days after discharge. It remained positive at the next two tests before March 5th, 2020 [fig_ref] Table 1: COVID-19 nucleic acid detection results of the cases 1 in different samples [/fig_ref].
## Case 4
A 38-year-old male was exposed when dining with his friends who had traveled to Wuhan. Three of the tables were infected with COVID-19. He presented with fever, fatigue, and cough for 8 days. He was diagnosed with novel coronavirus pneumonia on January 30, 2020, according to typical respiratory symptoms, positive oropharyngeal swab results and radiological ground-glass opacification. He was transferred to the designated hospital, and was treated for 27 days. The nasopharyngeal swab was positive 3 days after discharge, but became negative 2 days later on March 4th, 2020 [fig_ref] Table 2: COVID-19 nucleic acid detection results of the cases 2 in different samples [/fig_ref].
# Discussion
With the intensive surveillance measures that have been taken to prevent and control epidemics by the Chinese government, a good impact has been achieved on the spread and treatment of COVID-19 in Wuhan and the whole country. However, according to a report in JAMA, a follow-up of the recovered patients showed a recurrence of RT-PCR positivity in some patients [bib_ref] Positive RT-PCR test results in patients recovered from COVID-19, Lan [/bib_ref] [fig_ref] Table 3: COVID-19 nucleic acid detection results of the cases 3 in different samples [/fig_ref]. Why does this happen? In general, specimens such as alveolar lavage fluid deposited deep in the lungs have a higher positive rate than that of oropharyngeal swabsand turn negative more slowly. If the patient is discharged after two negative results of oropharyngeal swab test separated by at least 24 h, there may still be a persistent virus in the lung. When patient immunity declines, the RT-PCR test of oropharyngeal swabs may be positive again. Thus it is recommended that saliva should be collected if possible, and it has been emphasized that specimens of the lower respiratory tract (saliva or airway extract) provide more accurate RT-PCR tests. In addition, virus-free cells may be collected due to improper sites and insufficient depth [bib_ref] Cause analysis and treatment strategies of "recurrence" with novel coronavirus pneumonia (covid-19)..., Zhou [/bib_ref] , which will also cause "false negatives". This phenomenon may also be due to the biological characteristics of the novel coronavirus [bib_ref] The way to reduce the "false negative results" of 2019 novel coronavirus..., Zhang [/bib_ref].
Positive results of the anal swab nucleic acid test do not mean that there is live virus in the stool of the patient. Second, the removal rate of viral RNA in the stool of the patient is slower than that of the oropharyngeal swab. Therefore the duration of positive anal swabs or stool samples during recovery is relatively long [bib_ref] Persistence and clearance of viral RNA in 2019 novel coronavirus disease rehabilitation..., Ling [/bib_ref].
There are several limitations in our observation. Virus nucleic acids of different samples were tested in the two hospitals and at the CDC, and samples were collected by different medical professionals in the above mentioned settings. Furthermore, as we were facing to a new virus, different batches of kits from different manufacturers were used, which might increase bias.
[fig] Funding: This work is funded by the Fundamental Research Funds for the Central Universities at Chongqing University (Grant No. 2019CDY-GYB020); Basic Research and Frontier Exploration of Science and Technology Commission by Chongqing Municipality (Grant No. cstc2018jcyjAX0335) and Yuzhong District (Grant No. 20180156) awarded to Dr. Wuquan Deng. [/fig]
[table] Table 1: COVID-19 nucleic acid detection results of the cases 1 in different samples. [/table]
[table] Table 2: COVID-19 nucleic acid detection results of the cases 2 in different samples. [/table]
[table] Table 3: COVID-19 nucleic acid detection results of the cases 3 in different samples. [/table]
|
Long-Term Venovenous Connection for Extracorporeal Carbon Dioxide Removal (ECCO2R)–Numerical Investigation of the Connection to the Common Iliac Veins
Associate Editor David Steinman oversaw the review of this article.AbstractPurpose-Currently used cannulae for extracorporeal carbon dioxide removal (ECCO 2 R) are associated with complications such as thrombosis and distal limb ischemia, especially for long-term use. We hypothesize that the risk of these complications is reducible by attaching hemodynamically optimized grafts to the patient's vessels. In this study, as a first step towards a long-term stable ECCO 2 R connection, we investigated the feasibility of a venovenous connection to the common iliac veins. To ensure its applicability, the drainage of reinfused blood (recirculation) and high wall shear stress (WSS) must be avoided.Methods-A reference model was selected for computational fluid dynamics, on the basis of the analysis of imaging data. Initially, a sensitivity analysis regarding recirculation was conducted using as variables: blood flow, the distance of drainage and return to the iliocaval junction, as well as the diameter and position of the grafts. Subsequently, the connection was optimized regarding recirculation and the WSS was evaluated. We validated the simulations in a silicone model traversed by dyed fluid. Results-The simulations were in good agreement with the validation measurements (mean deviation 1.64%). The recirculation ranged from 32.1 to 0%. The maximum WSS did not exceed 5.57 Pa. The position and diameter of the return graft show the highest influence on recirculation. A correlation was ascertained between recirculation and WSS. Overall, an inflow jet directed at a vessel wall entails not only high WSS, but also a flow separation and thereby an increased recirculation. Therefore, return grafts aligned to the vena cava are crucial. Conclusion-In conclusion, a connection without recirculation could be feasible and therefore provides a promising option for a long-term ECCO 2 R connection.
# Introduction
Since its first successful application in 1971, [bib_ref] Prolonged extracorporeal oxygenation for acute post-traumatic respiratory failure (shock-lung syndrome), Hill [/bib_ref] extracorporeal life support (ECLS, at that time called extracorporeal membrane oxygenation (ECMO) [bib_ref] The extracorporeal life support organization maastricht treaty for nomenclature in extracorporeal life..., Conrad [/bib_ref] has become an established treatment for circulatory and pulmonary support, and a combination thereof. [bib_ref] Extra Corporeal Membrane Oxygenation (ECMO) review of a lifesaving technology, Makdisi [/bib_ref] ECLS was originally a modified cardiopulmonary bypass, the use of which was limited to short-term support only. The longer duration of support with ECLS was facilitated by the simplification of the circuit. Nonetheless, the development was mainly caused by the introduction of polymethylpentene (PMP) hollow fiber membranes. [bib_ref] Hollow fiber oxygenator composition has a significant impact on failure rates in..., Daniel [/bib_ref] Yet, patients were still sedated and intubated. In the last 10-15 years, the application of ECLS has evolved to a mid-term support with awake and spontaneously breathing patients treated in the ICU. [bib_ref] Current and future status of extracorporeal life support for respiratory failure in..., Bartlett [/bib_ref] Again, this development was supported by technological advances, specifically in the field of oxygenators and pumps, [bib_ref] Contemporary extracorporeal membrane oxygenation for adult respiratory failure: life support in the..., Maclaren [/bib_ref] and by novel double-lumen cannulae, making ECLS simpler and safer. [bib_ref] Current and future status of extracorporeal life support for respiratory failure in..., Bartlett [/bib_ref] As a result, the use of ECLS increased by 433% in the US from 2006 to 2011. [bib_ref] Extracorporeal membrane oxygenation use has increased by 433% in adults in the..., Sauer [/bib_ref] Today, ECLS patients can be awake and ambulatory [bib_ref] Current and future status of extracorporeal life support for respiratory failure in..., Bartlett [/bib_ref] and treated for prolonged time periods up to several months or occasionally even longer. [bib_ref] Late recovery from total lung injury after ECMO support, Bartlett [/bib_ref] [bib_ref] One hundred seven days of ECMO as a bridge to lung transplantation:..., Kim [/bib_ref] [bib_ref] Clinical outcomes of patients receiving prolonged extracorporeal membrane oxygenation for respiratory support, Na [/bib_ref] [bib_ref] One hundred ten days of extracorporeal membrane oxygenation in a young woman..., Strecker [/bib_ref] [bib_ref] Prolonged VV ECMO (265 Days) for ARDS without technical complications, Wiktor [/bib_ref] This enables the use of ECLS for new indications, for example as bridge to lung transplantation. [bib_ref] ECMO as bridge to lung transplant, Biscotti [/bib_ref] [bib_ref] Extracorporeal membrane oxygenation as a bridge to pulmonary transplantation, Hoopes [/bib_ref] [bib_ref] Successful lung transplant after prolonged extracorporeal membrane oxygenation (ECMO) in a child..., Tissot [/bib_ref] Furthermore, new extracorporeal applications have emerged, such as extracorporeal carbon dioxide removal (ECCO 2 R) In general, the ambulation of patients is associated with better outcomes at hospital discharge, [bib_ref] Early physical and occupational therapy in mechanically ventilated, critically ill patients: a..., Schweickert [/bib_ref] which has also been confirmed for ECLS. [bib_ref] Cannulation for veno-venous extracorporeal membrane oxygenation, Lindholm [/bib_ref] In the future, as postulated by [bib_ref] Current and future status of extracorporeal life support for respiratory failure in..., Bartlett [/bib_ref] ECLS will be wearable thus facilitating a management in step-down units, general care or even at home and increasing the quality of life for patients. Additionally, ECLS will be used as destination therapy, at first for the palliation of severe COPD. Yet, for the realization of this next milestone in ECLS therapy, further technological improvements are needed. Current research focuses on the development of wearable and integrated devices, such as the MLung, [bib_ref] A low resistance, concentric-gated pediatric artificial lung for end-stage lung failure, Thompson [/bib_ref] PAL [bib_ref] In vivo 5 day animal studies of a compact, wearable pumping artificial..., Madhani [/bib_ref] and RasQ. [bib_ref] In-vivo study of a novel longterm lung support device for PAH, bridge-to-transplant..., Borchardt [/bib_ref] However, these new integrated and wearable devices designed for ambulatory long-term treatment are in need of improved connections to the patient's vessels. [bib_ref] How computational modeling can help to predict gas transfer in artificial lungs..., Kaesler [/bib_ref] Currently used cannulae are associated with several complications and disadvantages, which prolong applications. These include cannula thrombosis, [bib_ref] Initial experience with single cannulation for venovenous extracorporeal oxygenation in adults, Bermudez [/bib_ref] limb ischemia, [bib_ref] The challenges faced with early mobilization of patients on extracorporeal membrane oxygenation, Javidfar [/bib_ref] [bib_ref] Cannulation strategies in adult veno-arterial and veno-venous extracorporeal membrane oxygenation: techniques, limitations,..., Jayaraman [/bib_ref] [bib_ref] Cannulation techniques for extracorporeal life support, Pavlushkov [/bib_ref] [bib_ref] Vascular access for extracorporeal life support: tips and tricks, Reeb [/bib_ref] [bib_ref] Factors associated with ipsilateral limb ischemia in patients undergoing femoral cannulation extracorporeal..., Yau [/bib_ref] risk of dislocation, [bib_ref] The challenges faced with early mobilization of patients on extracorporeal membrane oxygenation, Javidfar [/bib_ref] [bib_ref] Cannulation techniques for extracorporeal life support, Pavlushkov [/bib_ref] infection of the cannulation site, [bib_ref] Cannulation strategies in adult veno-arterial and veno-venous extracorporeal membrane oxygenation: techniques, limitations,..., Jayaraman [/bib_ref] [bib_ref] Cannulation techniques for extracorporeal life support, Pavlushkov [/bib_ref] limitation of patient's mobility, [bib_ref] The challenges faced with early mobilization of patients on extracorporeal membrane oxygenation, Javidfar [/bib_ref] [bib_ref] Cannulation strategies in adult veno-arterial and veno-venous extracorporeal membrane oxygenation: techniques, limitations,..., Jayaraman [/bib_ref] [bib_ref] Cannulation techniques for extracorporeal life support, Pavlushkov [/bib_ref] high effort of ambulation, [bib_ref] The challenges faced with early mobilization of patients on extracorporeal membrane oxygenation, Javidfar [/bib_ref] [bib_ref] Cannulation strategies in adult veno-arterial and veno-venous extracorporeal membrane oxygenation: techniques, limitations,..., Jayaraman [/bib_ref] [bib_ref] safety and feasibility of early physical therapy for patients on extracorporeal membrane..., Wells [/bib_ref] and arguably a low quality of life.
Therefore, we propose a new venovenous (VV) connection to the patient's common iliac vessels (CIV), which is implemented using grafts. The iliac anastomosis offers the possibility of an extraperitoneal approach in analogy to a kidney transplant. We hypothesize that this connection could provide sufficient blood flow for elective interventions for longterm lung support of chronic patients, such as EC-CO 2 R, without disturbing the physiological blood flow. The grafts could exit in the abdominal region without hindering the patient's mobility. When connecting to the iliac vessels, it is possible to guide the grafts to the exit site without puncturing the peritoneum. The connection is realized as an end-to-side anastomosis, a technique that has been proven to be long-term stable for connections of ventricular assist devices.
In this study, we investigate the connection of grafts to the CIVs for venovenous ECCO 2 R. Both drainage and return grafts are connected to a CIV on either side before the convergence to the inferior vena cava (IVC). Since both grafts are connected to the same blood vessel and in proximity of each other, the blood returned from the extracorporeal circulation (ECC) to the patient may be directly withdrawn back to the ECC by the drainage graft Since both grafts are connected to the same circulation and in proximity of each other, the returned blood may be directly withdrawn by the drainage graft. This process is called ''recirculation'' [bib_ref] Cannulation strategies in adult veno-arterial and veno-venous extracorporeal membrane oxygenation: techniques, limitations,..., Jayaraman [/bib_ref] [bib_ref] Vascular access for extracorporeal life support: tips and tricks, Reeb [/bib_ref] and is suspected to be a major issue of the proposed connection.
Recirculated blood through the ECC entails a decreased gas exchange efficiency, which is detrimental-especially in small devices for ambulatory applications. Therefore, it was selected as the primary focus of the study. Wall shear stress (WSS) was selected as the secondary focus of the study, as it allows conclusions to be drawn about the severity of the iatrogenic changes to the blood flow. Both factors were evaluated to deduce the possibility of a connection configuration that avoids recirculation and induces only minimal changes to the physiological blood flow.
In summary, we hypothesize that it is possible to connect the drainage and return of an ECC to the left and right CIV, respectively, using grafts, without recirculation and maintaining a physiological blood flow.
# Materials and methods
The establishment of an anatomical model was needed for both the numerical simulation of the flow characteristics of the novel ECCO 2 R connection and the subsequent in-vitro validation thereof.
## Anatomical model
To identify a suitable anatomical model for the site of connection, imaging data of patients in supine position were analyzed and evaluated. A board-certified radiologist selected CT data of 24 patients with healthy vascular systems. All imaging data were anonymized. Three data sets were excluded due to insufficient image quality, two data sets due to abnormal anatomy of the venous vessels, and one data set due to an extreme deformation of the spine. We segmented and measured the remaining 18 data sets at distinctive locations [fig_ref] FIGURE 1: Measurement locations with corresponding denominations in one exemplary anatomy [/fig_ref]. Therein, D1 is the diameter of the IVC, measured at the level of the first lumbar vertebra and D2, also the diameter of the IVC, but measured just after the convergence of the CIVs to the IVC. D3, D4, D5, and D6 are diameters of the CIVs: D3 (right branch) and D4 (left) are measured just before the convergence into the IVC whereas D5 (right) and D6 (left) are measured just after the convergence of the external and internal iliac vein into the CIVs. D7, D8, D9, and D10 are diameters of the external iliac vein, whereby D7 (right) and D8 (left) are measured just before the convergence of the external and internal iliac vein into the CIVs and D9 (right) and D10 (left) are measured just after the transition from the femoral vein to the external iliac vein. Additionally, we measured the distances between D1 and D2, D3 and D5 (length of right CIV), and D4 and D6 (length of left CIV). Furthermore, the angles between IVC and each CIV as well as the angle between both CIVs were measured. The segmentation and measurement of the diameters and distances were performed using Mimics Research 20.0 (Materialise NV, Leuven, Belgium). For the measurement of the angles, we exported the centerline into Creo Parametric 4.0 (PTC Inc., Boston, MA).
The arithmetic mean, median, standard deviation, median absolute deviation, and coefficient of variation were calculated for each measurement of a diameter, distance, and angle. To identify a suitable patient anatomy for the evaluation of the novel connection, the percentage deviation from the average value for each data set for each measurement was calculated. Thereafter, the average percentage deviation was calculated for each data set and the one with the lowest value was chosen as the most suitable anatomy. This anatomy was transferred into 3-matic Research 13.0 (Materialise NV, Leuven, Belgium), where it was smoothed and fixed to eliminate e.g. holes and overlaps.
## Computational fluid dynamics
The chosen anatomical model was supplemented by two vascular grafts (DesignModeler, Ansys 19.0, AN-SYS, Inc., Canonsburg, PA) in order to evaluate different connection configurations by numerical simulation: the drainage graft on the left CIV and the return graft on the right CIV. Both grafts were connected with an angle of 25°, as this was previously found to be beneficial and to entail near-physiological WSS. [bib_ref] The influence of the anastomosis angle on the hemodynamics in the distal..., Grus [/bib_ref] [bib_ref] The ratio of diameters between the target artery and the bypass modifies..., Grus [/bib_ref] We implemented the grafts as ideally straight tubes and assumed a perfect anastomosis with an oval orifice for all simulations. Three parameters were varied in the numerical simulations:
Diameter of the grafts Distance of the grafts to the iliocaval junction (IJ)
Position of the grafts, defined by the rotational angle around the native flow direction
The diameter of the grafts was varied between 7 and 9.5 mm. For the distance of the grafts to the IJ, we chose two discrete positions. One position was located close to the IJ, denoted as ''high'' and one was located more caudally, denoted as ''low''. Both the drainage and return positions were varied independently, which yielded four different setups. We were unable to implement a program controlled variation of the distance because we used a physiological anatomy, which is complex, skewed, and tortuous. Therefore, the location of the intersection of graft and CIV would vary, which led to the termination of the simulations either because of a faulty, or a failed connection of the CIV and the graft. In each of the four setups, the rotational angles were restricted either by the opposing CIV or graft and therefore, the limits had to be established individually.
The different setups and the corresponding distances and ranges of the rotational angles are shown in [fig_ref] TABLE 1: The different setups for the CFD simulations including the distance of the... [/fig_ref]. [fig_ref] FIGURE 2: Schematic presentation of the rotational angles of the grafts and the respective... [/fig_ref] provides a schematic depiction of the rotational angles of the grafts and the respective distances to the IJ.
The first part of the numerical study was designed to investigate the influence of the different parameters. Therefore, 20 design points were created for each of the chosen setups using Optimal-Space-Filling-Design (Ansys 19.0, ANSYS, Inc., Canonsburg, PA). This function facilitates the generation of sensibly distributed configurations for optimal parameter correlation. Subsequently, CFD simulations were performed for all created design points. Based on these simulations, a parameter response surface was created from which the influence of the different parameters on the recirculation can be derived. The second part of the numerical study was designed to determine the maximum and minimum values for the recirculation. For this purpose, we used the optimization function provided by Ansys and chose the screening method with 1,000 samples for each setup. The response surface is used to calculate additional candidate points without simulation and thus identifies graft configuration with minimum and maximum recirculation. The identified optimal candidate points and results were validated by additional CFD simulations and added to the overall design points for the evaluation. Additionally, the WSS on the vessel wall for each design point was calculated.
For this study, the return blood flow in the CIVs was set to 1.0 L/min each, as the lower margin of the physiological range. For the drainage blood flow at the IVC, a pressure boundary condition with 6.0 mmHg was chosen. We investigated two different scenarios, represented by different blood flows through the grafts: On the one hand, a critical but still acceptable scenario, in which the flow through the grafts is at the upper end of blood flows clinically used in ECCO 2 R (0.5-1.0 L/ min).This resulted in a graft flow of 100% of the CIV flow. On the other hand, we chose a worst-case scenario, in which the patient's demand for extracorporeal CO 2 removal increases either due to a deteriorating lung function or an increased demand caused by the ambulation. Consequently, the blood flow through the ECC is increased, which can entail a graft flow that exceeds the native blood flow. This resulted in a graft flow of 150% of the CIV flow. Additionally, we assumed that this amplifies the recirculation as the corresponding CIV does not provide enough blood for the drainage and thereby more returned blood is drawn to the drainage graft. Therefore, the influences of the varied parameters on the recirculation are augmented, providing more insight into the effects. To allow a reasonable comparison to the validation experiments, blood was implemented as a Newtonian fluid with a dynamic viscosity of 3.6 mPas and a density of 1056.4 kg/m 3 , which corresponds to a hematocrit of 44%. For the investigation of recirculation two fluids with identical properties, as described above, were created. One fluid enters through both CIVs and the other one enters at the return graft. Thus, the recirculation is calculated as the ratio of the mass flow of the second fluid to the total mass flow through the drainage graft. All numerical simulations were performed steady state using Ansys CFX 19.0 (ANSYS, Inc., Canonsburg, PA). The RNG-j-e-Model was used to model turbulence. The geometry was meshed using Ansys Meshing 19.0 (ANSYS, Inc., Canonsburg, PA). The mesh parameters were set to: 5 inflation layers at the vessel and graft wall, growth rate of 1.2, and first layer thickness of 0.1 mm, to achieve a dimensionless distance from the wall of below 2 (y + < 2). The global element sizing (0.3-0.8 mm) was varied to study the mesh sensitivity. This method was taken from Thamsen et al. [bib_ref] Numerical analysis of blood damage potential of the HeartMate II and HeartWare..., Thamsen [/bib_ref] Six meshes with different numbers of elements were compared regarding the results of the recirculation. Since this study focused on the investigation of a broad range of configurations, and with regard to a feasible computational time, a 5% deviation of the recirculation to the finest mesh was deemed acceptable in accordance with the literature (acceptance values 3-9% 2,15 ). Thus, mesh 4 was chosen for the further analyses and additionally, an experimental validation was performed to ensure the validity of the results. The characteristics of the meshes, the resulting recirculations, and the percentage deviation of each recirculation to the result of the finest mesh are shown in [fig_ref] TABLE 2: Characteristics, resulting recirculations, and percentage deviation of each recirculation to the result... [/fig_ref]. [fig_ref] FIGURE 3: Recirculation dependent on the number of the elements of the 6 different... [/fig_ref] depicts the recirculation dependent on the number of elements of the meshes included in the sensitivity study.
## Validation experiments
With one setup and both numerically investigated graft flows we performed dye experiments to validate our CFD simulations. We chose the setup Rh_Dl with a favorable position of the grafts for manufacturing and a high recirculation. The geometry was manufactured by rapid prototyping (Objet 350 Connex3, Stratasys, Ltd., Eden Prairie, MN) of a liquid acrylicbased photopolymer (VeroClear TM , Stratasys). The support material was removed mechanically first roughly by hand and then chemically in a 2% sodium hydroxide solution. Afterwards, the geometry was sanded to obtain a smooth surface. The manufactured geometry was then placed in a box, which was subsequently filled with transparent silicone (ELASTOSILÒ RT 601 A/B, Wacker Chemie AG, Munich, Germany). At the top end of the IVC, a 1/8'' PVC tubing (T) was placed perpendicular to the flow direction to function as a manometer. The geometry was connected to two different flow loops [fig_ref] FIGURE 4: Experimental setup for the validation of the CFD simulations [/fig_ref] Immediately before the experiments, we determined a calibration curve for the color sensor, by measuring different mixtures of clear and dyed fluid. The mixtures were 100, , and 0% of dyed fluid. Additionally, the sensitivity of the color sensor was assessed by incrementally adding dye to the clear fluid and recording the resulting signal. The first ratio that produced a sensor signal was defined as the minimum recirculation measurable. At the start of the experiments the loop was filled with clear fluid and the threeway valves were connected to each other. With this configuration, the flows and pressures were adjusted to match the values used in the CFD simulation: 2.0 L/ min IVC flow, which results in 1.0 L/min for each CIV, 1.5 or 1.0 L/min graft flow, depending on the experiment, and a pressure of 6 mmHg at the top of the IVC. Subsequently, both three-way valves were switched simultaneously, whereby the dyed fluid was added to the loop, flowed through the return graft, and mixed with the clear fluid from the CIVs. The resulting fluid was measured with the color sensor and discarded. Every experiment was repeated five times.
# Results
## Anatomical model
The results of the anatomical analysis of the 18 image data sets are presented in [fig_ref] FIGURE 5: Boxplots of the anatomical measurements, wherein the value of the selected anatomy... [/fig_ref]. The mean diameter ± standard deviation of the IVC at the level of the first lumbar vertebra is 25.8 ± 3.3 mm and decreases on a mean length of 138.4 ± 17.8 mm towards the CIVs to a mean diameter of 22.9 ± 2.7 mm. Just before the convergence of the CIVs to the IVC, the right and the left CIV have a mean diameter of 15.9 mm with a standard deviation of 2.8 mm and 3.3 mm, respectively. The right CIV is on average 63.6 ± 18.0 mm long, whereas the left CIV is longer, with an average length of 81.4 ± 14.8 mm. The right and left vena iliaca externa show similar diameters just before the convergence into each CIV, 14.4 ± 2.4 mm and 14.3 ± 2.9 mm, respectively. The same applies for the diameters at the beginning of both right and left vena iliaca externa, as they show a diameter of 13.3 ± 2.6 mm and 13.1 ± 2.4 mm, respectively. The angles between the IVC and each CIV show the highest variances: The angle between IVC and right and left CIV are 142.0 ± 37.1°and 135.7 ± 37.4°, respectively. The angle between both CIVs shows a lower variance, with 83.3 ± 6.1°.
The performed analysis entailed two candidate anatomies for the subsequent CFD simulations: subject 3 and 10. Because of the limited field of view of subject 10, a measurement of D1 was not possible and therefore, we selected the anatomy of subject 3. The respective measurements of subject 3 are marked in [fig_ref] FIGURE 5: Boxplots of the anatomical measurements, wherein the value of the selected anatomy... [/fig_ref]. [fig_ref] FIGURE 6: Streamlines and WSS contour plots of the highest [/fig_ref] depicts the streamlines of the highest and lowest recirculation and contour plots of the WSS for each graft configuration. The graft flows of the worstcase scenario of 150% of the CIV flows leads to recirculations of up to 32.3%. In comparison, the graft flow of the critical scenario entails overall significantly lower recirculations with a maximum of 8.9%. The WSS contour plots show two areas prone to slightly elevated WSS on the vessel wall: one around the return graft and another one where an inflow jet hits the vessel wall. These areas of impingement seem to be more pronounced in the graft configurations with the maximum recirculation. Also apparent is an increase of the maximum WSS for the 150% graft flow compared to the 100% graft flow.
## Computational fluid dynamics
The results of the sensitivity study for a graft flow of 150% identify the rotational angle of the return graft as the parameter that affects the recirculation most, with a mean value of 0.50, followed by the diameter of the return graft with 0.21. For a graft flow of 100%, the recirculation is most sensitive to the diameter of the return graft, with a value of 0.51, followed by the rotational angle of the inlet graft with 0.43. Additionally, a weak correlation between an increased recirculation and an increased WSS could be derived from [fig_ref] FIGURE 7: Scatter plots of maximum WSS dependent on recirculation [/fig_ref]. Also apparent in this figure, for a 150% graft flow, is a connection between the return graft in the lower location and an increased recirculation: The lower location of the return graft results in recirculations of more than 30%, whereas the higher location of the return graft entails maximum recirculations of mostly less than 20%. The location of the drainage graft, on the other hand, does not seem to have an influence on the recirculation. Contrary to this, for a 100% graft flow, the drainage graft location affects the amount of recirculation: The drainage graft in the higher location results in recirculations of more than 6%, whereas the drainage graft in the lower position entails recirculations of less than 3%. Here, the location of the return graft does not appear to impact the recirculation. The maximum WSS for the simulations with 150 and 100% graft flow did not exceed 5.57 Pa and 2.41 Pa, respectively. Hence, the decrease of the graft flow by 33%, decreases the maximum WSS by approximately 60%. Scatter plots of the recirculation dependent on the varied parameters (angle of drainage and return graft, as well as diameter of drainage and return graft) are provided in the appendix (Figs. [bib_ref] Modeling of non-Newtonian blood flow through a stenosed artery incorporating fluid-structure interaction, Chan [/bib_ref] [bib_ref] Common iliac vein stenosis and risk of symptomatic pulmonary embolism: an inverse..., Chan [/bib_ref] [bib_ref] The extracorporeal life support organization maastricht treaty for nomenclature in extracorporeal life..., Conrad [/bib_ref].
## Biomedical engineering society
## Experimental validation
The calculated values for the recirculation from the CFD simulations are shown in [fig_ref] FIGURE 2: Schematic presentation of the rotational angles of the grafts and the respective... [/fig_ref].04 and 1.0% for a graft flow of 1.5 and 1.0 L/min, respectively. The figure also depicts the calibration curves of the four color sensor channels. On the curves, the measured sensor signals in the five experiments for a graft flow of 1.5 L/ min are matched with the corresponding recirculation. The measured values were 232.4 ± 16.8 (red), 334.3 ± 21.3 (green), 503.3 ± 22.6 (blue), and 998.3 ± 57.5 (clear). Converted into percentage recirculation using the calibration curve, these values correspond to 23.5 ± 2.4% (red), 23.3 ± 2.2% (green), 23.8 ± 2.3% (blue), and 23.9 ± 2.4%. The recirculation of a graft flow of 1.0 L/ min did not yield any measurement values, because the sensitivity of the color sensor only allowed measurements of a minimum of 1.5% recirculation. A qualitative comparison of the CFD simulations with the experimental validation is shown in [fig_ref] FIGURE 9: Qualitative comparison of CFD simulations and validation experiments for the configuration Rh_Dl... [/fig_ref].
# Discussion
In the present study, we investigated the possibility of connecting an ECC for ECCO 2 R to the common iliac veins using grafts. First, the anatomy of the iliac veins was analyzed and an anatomical model for the subsequent CFD simulations was chosen. Second, the diameter of the grafts, the rotational angle around the native flow direction, and the height of the grafts were varied in the CFD simulations using the developed anatomical model, before an optimization regarding recirculation was performed. Third, we validated the CFD simulation experimentally in a silicone model of the iliac and cava veins by conducting dye tests. The results of the anatomical measurements show relatively low variations, except for the angles between the CIVs and the IVC. Comparing the diameters along the progression of the IVC into the CIVs, and farther on the external iliac veins, entails decreasing values. Therefore, we consider the anatomical study as reasonable. However, our results are at the upper end of the values reported in the literature for CIV diameters: e.g. 11.2 ± 3.3 mm, [bib_ref] Common iliac vein stenosis and risk of symptomatic pulmonary embolism: an inverse..., Chan [/bib_ref] 11.5 mm (range 8.7-13.9 mm), [bib_ref] Optimal sizing of iliac vein stents, Raju [/bib_ref] and 11.5 mm (range 6.3-16.1 mm). [bib_ref] Computed tomography findings in 10 cases of iliac vein compression (May-Thurner) syndrome, Oguzkurt [/bib_ref] Additionally, our selected anatomy has a relatively small diameter of the right CIV and the angles between IVC and CIVs are only just inside the interquartile range. Nevertheless, this anatomy showed the lowest overall variance, which may allow a more general interpretation of the effects of the parameters, regardless of the patientspecific geometry.
The experimental validation for a 150% graft flow is in good agreement with the CFD simulations, with a mean deviation across all measurements and channels of 1.64%. This is corroborated by the experiments for the 100% graft flow: No recirculation was measurable, because the expected recirculation of 1.0% is less than the sensitivity of the color sensor, which is 1.5%. In addition, the qualitative comparison of the flow patterns (see [fig_ref] FIGURE 9: Qualitative comparison of CFD simulations and validation experiments for the configuration Rh_Dl... [/fig_ref] , shows a large similarity. However, for the 100% graft flow experiments, there is a noticeable difference above the return graft: The re-sults of the CFD simulations show an area on the left side of the IVC without any fluid from the return graft. On the contrary, the image of the experiments shows this area filled with dye. Therefore, fluid from the return graft must have entered this area. We assume that this happens due to transient effects induced by the switching of the two three-way valves at the start of the dye measurements.
## Biomedical engineering society
Traditionally, particle image velocimetry (PIV) is used as a validation experiment for CFD simulations. With PIV the flow field can be visualized and the flow velocity can be measured. From this, the WSS can be determined, which is then compared with the simulations. However, PIV does not allow a distinction of different fluids and therefore a quantification of the recirculation is not feasible. Thus, we chose the described dye experiments to validate our simulations, because the recirculation was the primary focus of the study. However, the data from these experiments cannot be used to validate the simulated WSS results.
For further experiments including a more detailed investigation of the WSS, PIV measurements should be chosen as additional validation experiments.
The reason for the predominant influence of the rotational angle of the return graft on the recirculation for a 150% graft flow is most likely related to ''jet splitting'': If the inflow jet is directed at the opposing wall of the IVC, the impact on the wall splits the jet. One part flows upwards, in the native flow direction of the IVC. The other part of the jet is directed towards the left CIV and the drainage graft. Thereby, the reinfused fluid can reach the drainage graft and is subsequently transported back through the ECC. The degree of recirculation is then dependent on the magnitude of the impingement of the jet on the vessel wall. This assumption is supported by the correlation of the recirculation and the average WSS, because a stronger impingement of the jet on the wall entails a higher WSS in the impingement region.
Considering these aspects, we deduce that the strong influence of the rotational angle on the recirculation is directly connected to the strong influence of the impingement angle on the recirculation, because the impingement angle itself depends on the rotational angle of the return graft. Differently rotated grafts direct the inflow jet either more towards the wall of the IVC or in the native flow direction of the IVC. This can also explain the increased recirculation of the configurations with the return graft in the lower location: There are supposedly more positions in the lower location that are prone to the inflow jet hitting the vessel wall. The higher locations are less prone to the inflow jet hitting the vessel wall, because the angle of the CIV to the IVC affects this position less. The sensitivity of the recirculation to the return graft diameter is likely due to decreased velocities of the inflow jet with larger diameters: The momentum of the inflow jet is decreased, which is why it can be more easily deflected by the native blood flow towards the IVC.
For the 100% graft flow simulation, the order of the sensitivities is reverse, with the rotational angle showing less influence on the recirculation. This can also be explained by the momentum of the inflow jet. The momentum is decreased even further, again allowing a better deflection of the inflow jet by the vessel flow. This is again supported by the WSS results, as they are lower as well.
Ultimately, for a 100% graft flow, there are more rotational angles of the return graft, in which the jet split is reduced or even avoided completely, than for a 150% graft flow. Consequently, the impact of the diameter of the return graft on the flow velocities becomes more notable. Also notable is the increased recirculation with a higher location of the drainage graft, which we also attribute to the decreased flow velocities. Hence, the resulting decreased momentum of the inflow jet results in a weaker jet split, causing a lower momentum of the jet directed towards the drainage graft. Therefore, the drainage graft in the higher location, closer to the impingement region, is reached more easily by the flow from the return graft before it can be redirected by the native flow from the CIV.
The comparison of the WSS contour plots of the highest and lowest recirculation in [fig_ref] FIGURE 6: Streamlines and WSS contour plots of the highest [/fig_ref] supports our hypothesis described above, as the graft configurations with maximum recirculation are associated with more pronounced areas of impingement. Nonetheless, [fig_ref] FIGURE 7: Scatter plots of maximum WSS dependent on recirculation [/fig_ref] indicates merely a weak correlation between WSS and recirculation. This could be attributed to the conflicting influence of the diameter of the return graft. Since one area of high WSS is located around the return graft, the maximum WSS in this area would increase with smaller graft diameters, as this induces higher velocities compared to grafts with larger diameters. Therefore, combinations of a small graft diameter and an angle of the return graft in a position that entails a low recirculation could account for graft configurations that weaken the correlation between WSS and recirculation.
There are several limitations of this study. (1) Even though we selected a reference model with the lowest deviation from the mean, our study was performed only for this specific anatomy. Therefore, generalizations of our conclusions should be treated conservatively. (2) Our study includes several simplifications regarding the graft connections, as they are implemented in simplified form as ideally straight tubes with a constant angle of 25°. More realistic graft designs, for example tapered, curved, or hooded grafts, as well as changing angles due to the movement of the patient, were not considered. These aspects can critically impact the flow behavior and thus, the recirculation and the WSS, which limits the generalization of the results. Additionally, our study was performed in a model with rigid walls and without dynamic changes in flow or pressure. For a long-term application with ambulated patients, these aspects should be considered in subsequent studies, because events like the collapsing of a graft might exclude the proposed connection from a clinical application. (3) The study is further limited by modeling the blood flow as a Newtonian fluid. In reality, blood is a non-Newtonian fluid and shows shear-thinning behavior with increasing shear rates. [bib_ref] Blood rheology and hemodynamics, Baskurt [/bib_ref] However, we chose the Newtonian model to allow a reasonable comparison to the validation experiments as there are no standard non-Newtonian blood-analogue fluids. Furthermore, the appropriate application of the various non-Newtonian models is subject to discussion in the literature [bib_ref] Modeling of non-Newtonian blood flow through a stenosed artery incorporating fluid-structure interaction, Chan [/bib_ref] [bib_ref] A computational fluid-structure interaction model of the blood flow in the healthy..., Razaghi [/bib_ref] and several studies confirm that Newtonian rheology provides a reasonable prediction for blood flow behavior. [bib_ref] Modeling of non-Newtonian blood flow through a stenosed artery incorporating fluid-structure interaction, Chan [/bib_ref] [bib_ref] Effect of rheological models on the hemodynamics within human aorta: CFD study..., Karimi [/bib_ref] [bib_ref] Study of Newtonian and non-Newtonian effect of blood flow in portal vein..., Khan [/bib_ref] [bib_ref] Effect of non-Newtonian and pulsatile blood flow on mass transport in the..., Liu [/bib_ref] [bib_ref] Newtonian and non-Newtonian blood flow in coiled cerebral aneurysms, Morales [/bib_ref] [bib_ref] Numerical simulation of blood pulsatile flow in a stenosed carotid artery using..., Razavi [/bib_ref] (4) Only the grafts themselves or the veins limited the positioning of the grafts; no other anatomies were taken into account, especially not of the traversing right common iliac artery. Therefore, it is possible that the resulting ideal graft configurations cannot be surgically applied in a clinical scenario. (5) A compromise between mesh size and computing time had to be found. As quantitative WSS evaluations require extremely fine meshes towards the wall, we decided to limit the evaluation of WSS values in this study to a qualitative comparison of the configurations. For quantitative assessment of the WSS, a refined wall mesh as well as the consideration of WSS in the mesh sensitivity study are necessary.
In conclusion, our study shows that connecting a VV ECCO 2 R to the CIVs using grafts may be possible without recirculation. Especially for a graft flow of 100% CIV flow the majority of graft configurations entail a negligible recirculation. Thus, optimum positioning of the grafts can completely avoid recirculation: The return graft should be aligned with the native flow direction of the IVC so that the inflow jet does not hit a vessel wall. Additionally, a larger graft diameter is preferable, both because of lower WSS and because of a lower risk of recirculation. Lastly, a location of the drainage graft farther away from the IVC is favorable. The angle and diameter of the drainage graft show a negligible influence on the recirculation. Especially for even smaller ratios of graft and CIV flows, as used for ECCO 2 R, [bib_ref] Physiological and technical considerations of extracorporeal CO 2 removal, Karagiannidis [/bib_ref] we assume that recirculation becomes even less likely.
The results show that this new connection has a potential for further investigation. The next step should be a more detailed analysis of few anatomically feasible graft configurations, including pulsatile flow and flexible vessel walls. Additionally, experiments with flexible silicone tubes, pulsatile flow, and realistic designs of the grafts could be performed. Ultimately, the feasibility of the new connection will have to be investigated in in-vivo experiments.
# Appendix
See Figs. 10, 11, 12, and 13.
[fig] FIGURE 1: Measurement locations with corresponding denominations in one exemplary anatomy. BIOMEDICAL ENGINEERING SOCIETY [/fig]
[fig] FIGURE 2: Schematic presentation of the rotational angles of the grafts and the respective distances to the IJ. [/fig]
[fig] FIGURE 4: Experimental setup for the validation of the CFD simulations. [/fig]
[fig] FIGURE 3: Recirculation dependent on the number of the elements of the 6 different meshes, used in the sensitivity study. The triangle represents the mesh, which we chose for the further analyses. [/fig]
[fig] FIGURE 5: Boxplots of the anatomical measurements, wherein the value of the selected anatomy are marked with a red ''X'': (a) Diameter D1, (b) Diameter D2, (c) Distance D1-D2, (d) Diameter D3 & D4, (e) Diameter D5 & D6, (f) Distances D3-D5 & D4-D6, (g) Angle G-L & G-R (h) Diameter D7 & D8, (i) Diameter D9 & D10. [/fig]
[fig] FIGURE 6: Streamlines and WSS contour plots of the highest (right) and lowest (left) recirculation for each graft configuration: (a)-(d) 100% graft flow, (e)-(h) 150% graft flow. [/fig]
[fig] FIGURE 7: Scatter plots of maximum WSS dependent on recirculation: (a)-(d) 100% graft flow, (e)-(h) 150% graft flow. [/fig]
[fig] FIGURE 9: Qualitative comparison of CFD simulations and validation experiments for the configuration Rh_Dl and a graft flow of 1.0 L/min (a) and 1.5 L/min (b). [/fig]
[fig] FIGURE 8: Calibration curves of the color sensor for the four channels with expected recirculation values calculated from CFD simulations (dashed vertical lines) and the results of the experimental validation (crosses). [/fig]
[table] TABLE 1: The different setups for the CFD simulations including the distance of the grafts to the IJ and the range of the rotational angle.The distance to the IJ was measured from lowest point of the IJ to the furthest intersection point of CIV and graft.. [/table]
[table] TABLE 2: Characteristics, resulting recirculations, and percentage deviation of each recirculation to the result of the finest mesh of the different meshes used for the mesh sensitivity study. [/table]
|
Density or Connectivity: What Are the Main Causes of the Spatial Proliferation of COVID-19 in Korea?
# Introduction
Since the outbreak of COVID-19 in December 2020, the virus has rampaged across the globe, causing 2.6 million deaths at the time of writing. South Korea has recorded a cumulated total of 92,817 positive cases and 1642 deaths. From its inception in China, the outbreak of the virus proliferated through Europe, the U.S., and eventually across all continents, at which point the World Health Organization officially declared it to be a pandemic. The sheer speed of the virus infection and the consequent heavy death toll strike fear in all people around the globe. Based on previous research on infectious diseases, such as AIDS, SARS, and Swine flu (the 2009 H1N1 Pandemic), the key determinants of virus spread were found to be closely related to the movement of people, urbanization, and the mass inflow of foreign nationals from overseas, all of which are characteristics often found in large metropolitan cities.
While large cities offer many benefits, tightly connected networks of goods and people around urban cores could accelerate the COVID-19 pandemic. Matthew and McDonaldnoted that the factors that precipitated the spread of malaria were economic development, patterns of land use, movement of people, and urbanization, whilst the spread of SARS and MERS were impacted by the inflow and outflow of people brought on by accelerated globalization. Other studies have also shown that transmission of infectious diseases is related to urban structure and people's movement between cities. For example, SARS, a viral respiratory illness, first originated from a wild animal market in Guangdong Province but later spread via international air travel into the world's cities. The 2014 West African Ebola virus shares similarities to SARS. Even though 64% and 60% of the population of Guinea and Sierra Leone lived in low-density agricultural areas, due to high connectivity and ease of travel, the two nations became the epicenter of the disease.
At first glance, places with a high degree of urbanization, such as New York, London, and Seoul, appear to be more susceptible to the proliferation of COVID-19. Previous studies have reported that infectious global diseases often leave behind a greater number of infections and a higher death toll. found that the 2009 H1N1 virus lasted longer in areas with a higher population density within Taiwan. Garrettalso noted that there existed a statistically significant relationship between the mortality rate of the 1918 Spanish Flu (influenza) and the population density per state. Thus, the concentration of population in a given area seems to be a catalyst for spreading infectious diseases. On the other hand, Nishiura et al.found that the 1918 Influenza in the UK and Japan showed no statistically significant relationship between population density, the severity of infection, and the mortality rate. Moreover, Parmet et al.found that during the 1918 Influenza, areas of lower density saw higher mortality rates than that of high-density cities.
Given these mixed findings, this study aimed to investigate the role of urban spatial structure on COVID-19 during the first eight months of the outbreak in Korea. Specifically, we focused on two factors related to the built environment: density and connectivity. By comparing different measures of density and connectivity, we sought to provide empirical evidence to determine which of the two measures are more related to the spread of COVID-19 in Korea. This research helps inform public health agencies and municipalities to find more effective measures of prevention and response against COVID-19 and other infectious disease threats.
## Literature review
## Cities and pandemics
Globalization and urbanization have progressed rapidly over the past 100 years. Globalization, prompted by the development of transportation and communication technologies, has been recognized as a distinct shift in the spatial and temporal dimensions of social and economic life, a phenomenon known as "time-space compression". This transition to the highly connected global marketplace has not only facilitated the movement of capital but also people and information, overcoming both distance and temporal obstacles of physical movement. However, globalization has not only positive effects of opening up new opportunities and new markets but also negative effects of promoting imbalances between countries, regions, or social classes. Moreover, globalization has led to an unexpected spread of disease by intensifying trade and human movement between countries, resulting in more frequent outbreaks of infectious diseases and their spread on a global scale. Before the rapid progress of globalization and urbanization, infectious diseases were more localized, and their spread was slow and intermittent.
Several high-profile infectious disease outbreaks occurred before the COVID-19 pandemic. For example, the 1918-1919 Influenza pandemic, known as the Spanish Flu, was the most severe pandemic in recent history, resulting in about 50 million deaths around the world. During this pandemic, the United States had implemented stringent social distancing measures, such as the closure of theaters, churches, schools, and saloons as well as the mandatory wearing of face masks. Since then, with the exception of the 1968 Hong Kong influenza pandemic, the outbreak and spread of infectious diseases were limited to certain regions or countries and were contained before spreading around the world. However, the current COVID-19, although often compared to the 1918 Influenza pandemic, is not just the return of the typical pandemic. This pandemic marks a new chapter in the history of infectious disease, demonstrating how quickly viruses can spread around the world in the 21st century with so many countries that are tied together in the global marketplace.
In both popular media and scholarly communities, there is an increasing debate around urban density and COVID-19 pandemic. On the one hand, cities are often characterized by their high-density urban form, which could be a risk factor for infectious disease outbreaks, such as COVID-19. Additionally, urban activities are not confined within the city boundary or a certain spatial domain because cities are agglomeration centers for socioeconomic connections of people, goods, and services. Higher connectivity within a city or between cities can amplify the risk of infectious diseases. On the other hand, cities can also provide a critical line of defense against pandemics. Responses to disease outbreaks could be faster and more efficient in larger cities with more resources and disaster preparedness. Cities have more diverse and higher quality health care systems, offering more choices and potential for their citizens to receive better care in times of crisis.
As we are responding to the immediate COVID-19 crisis, this pandemic highlights the need to start thinking about how to rebuild our cities and neighborhoods against future shocks. For example, urban planners have been increasingly discussing alternative urban models, such as "Superblocks", "Tactical urbanism", or "15-min City" to overcome some of the major problems that have been exposed during the pandemic. Before we start thinking about the future of cities after the pandemic, we first need to come to the drawing board and make a good assessment of which characteristics of cities are more vulnerable to pandemics and how we might be able to address them.
## Density and connectivity
Many empirical studies have attempted to assess the effects of urban factors on the spread of infectious disease, but the results have been mixed. In a study of the 1918 Spanish Influenza (H1N1), Garrettreported that there was a statistically significant relationship between mortality and population density and claimed that high-populationdensity counties served as a catalyst that accelerated the spread of the infectious disease. Arbel et al.assessed the influence of population density and socioeconomic factors on COVID-19 infection rates in Israel and found that the possibility of COVID-19 infection grew with population density levels. Ehlertattempted to find the relationship between density and socio-economic variables vis-a-vis mortality rate. He reported that population density and employment density were positively correlated with COVID-19 mortality rates. Wand and Lifound that population density mattered in the spreading of COVID-19 on the U.S. county level. They demonstrated that population density alone accounted for up to 76 percent of cumulative infection cases in the U.S. from early March to late May 2020. However, Agnoletti et al., reported that density was not significantly associated with the distribution of COVID-19 in Italy. After reviewing the most up-todate state of knowledge, UN-Habitatreported that density in itself appears not to be a decisive factor in the spread of the virus; but rather, inequality around income, race, service provision, and pre-existing health conditions seemed to have played a critical role in aggravating the vulnerability of individuals.
Another important factor related to urban spatial structures concerns connectivity. In a study of the 1918 Spanish Influenza in England and Wales, Chowellexplored the association between influenza death rates and a number of geographical and demographical indicators. While they found that cities had 30-40% higher death rates than rural areas, there was little association between mortality and measures of population density or residential crowding. The World Health Organizationreported that even though the majority of the population of Guinea and Sierra Leone live in rural areas, high cross-border connectivity due to the ease of travel seemed to have caused the wide spread of Ebola, making the two nations the epicenter of the pandemic. In a recent study of the COVID-19 pandemic in the U.S., Hamidi et al.found that urban density was unrelated to COVID-19 infection rates and even inversely related to death rates. They reported that connectivity between counties had a greater impact on both the infection and mortality rates than the population density of a given county. Similarly, reviewing research papers on the impact of urban density on COVID-19, Tellerargued that its impact was not closely related to density but connectivity.
## Research gap and our contribution
While there is a growing body of research examining the impact of urban spatial structure on COVID-19 outcomes, empirical evidence is mixed and inconclusive. These mixed results may stem from the use of different measurements across the study. Especially, the results could differ by how the dependent variable is measured and calculated. For example, in research conducted by Hamidi et al., when the number of total confirmed COVID-19 cases was used as the dependent variable, the results for population density were found to be statistically significant. However, when the dependent variable was redefined as the number of confirmed cases per 10,000 population, the result was statistically insignificant. Therefore, it is important to assess how robust the results are to different definitions of the dependent variable.
Moreover, some measures of urban densities could be biased when the denominator includes undevelopable lands, such as mountain ranges and rivers. Hence, we should take caution while calculating population and employment densities by using administrative regions that only include inhabitable areas. The measure of connectivity could also affect study results. For example, Hamidi et al.used a proxy measure of connectivity computed as a ratio of yearly airplane passengers per 10,000 population. While this is an appropriate measure in large countries with many domestic airline services, such measure may not be appropriate in smaller countries with limited airline travels. In addition, Tolleridentified that only a few empirical studies measured connectivity to identify diffusion of COVID-19.
In assessing the impact of connectivity, the social network analysis has been used in the infectious diseases studies to identify geographic transmission hubs of the infection as well as geographic distribution of infected patients. In addition, the analysis could be more useful for representing a holistic structure of the urban transportation network. The indicators measured using their analysis method allowed for examining the degree of connection between nodes at the specified spatial unit which, in this case, are cities. Cho et al.used the social network theory to analyze the characteristics of Chinese railway networks that have clear points of departure and arrival. The data on the points of departure and arrival for travel were used to compute mobility variation between cities. Shinalso analyzed the effects of COVID-19 on South East Asian airports' network centrality and found that the density and centrality of the airport network have shrunk during the initial phase of the COVID-19 pandemic. We followed these social network methods to develop the indicators of degree centrality, closeness centrality, betweenness centrality, and eigenvector centrality.
In this study, we aimed to fill the gap in the literature by comparing different measures of urban density and connectivity during the first eight months of the outbreak in Korea. We make contributions to the existing literature on three perspectives. First, we used the actual density indicators that take into account only habitable lands. Second, we used the aggregated origination and destination (O/D) data for all travel modes to calculate the degree of connectivity, closeness centrality, betweenness centrality, and eigenvector centrality to gain insight into how the connectivity of the overall transportation network affected the spread of COVID-19. Finally, we tested the sensitivity of the results by using different measures of dependent variables. This study provides insights into multiple factors of urban spatial structure, reflecting the current debate at play on the vulnerability of densely populated cities.
# Materials and methods
## Study area
Our study area was the entire country of South Korea to empirically identify the main causes of the spatial proliferation of COVID-19. In response to the COVID-19 outbreak, the Korean government has adopted a stringent contact-tracing strategy, called the 3T framework (Testing-Tracing-Treatment), to control and reduce the number of COVID-19 cases. Because of the effective control strategies and low infection rates, the Korean government has not imposed any national lockdown policies commonly adopted in the U.S. and most European countries. Soon after the COVID-19 outbreak in Korea, the health officials swiftly installed drive-and walk-through testing facilities, adopted extensive movement-and contract-tracing strategies, and communicated movement trajectories of confirmed patients via online maps and automatic mobile phone alert systems. Most Korean citizens actively followed the quarantine guidelines, such as mask-wearing and hand-washing, while voluntarily practicing strong social distance measures in highincidence areas as a self-protection measure.
Moreover, private firms as well as public offices implemented flexible work schedules, such as working from home and flexible working hours, as well as encouraging their employees to hold virtual meetings in lieu of in-person meetings. A survey conducted by the Korea Enterprises Federation (KEF)among the top 100 sales firms in September 2020 revealed that about 88 percent of the responded firms have implemented some form of work-from-home measures. Firms have adopted a wide range of work-from-home strategies to adapt to the rapidly escalating situation while continuing their operations remotely. Most of these behavioral and institutional changes were made possible relatively smoothly, largely owning to the past experience of the Middle East respiratory syndrome (MERS) outbreak, which caused great damage to the nation's economy back in 2012. Korea is considered one of a few countries that has prevented the rapid spread of COVID-19 without implementing the national lockdown widely adopted by other high-income countries, while also sustaining the national health and economic systems.
Thanks to the Korean government's rapid response and strict hygiene practices adopted by citizens and private firms, the infection rates have stayed well under control, with the rates ranging between 2 and 2.5% and even staying at 3.3% during the third spike in December 2020. This level is well below the 10% benchmark level that the World Health Organization recommends as adequate test-positivity levels. In addition, a recent review of 29 COVID-19 studies reporting reproductive numbersindicated that Korea was the only country that reported less than 1.0 in its estimated reproduction number (R, 0.76, 95% CI, 0.34-1.70). Other countries showed much higher reproductive numbers, reporting 3.14 for mainland China followed by 5.08 for Spain, 6.07 for Germany, and 6.32 for France.
The spatial unit of analysis in our study was a city, a county, and a district level, which is a smaller geographic unit than counties in the U.S.and in Germany. Several studies of COVID-19 have used similar geographic units (e.g., counties) to capture the effects of population density and connectivity on the spread of COVID-19. Moreover, the data on the number of confirmed cases for COVID-19 were only attainable at the city, county, and district levels, which are named Si, Gun, and Gu in Korean, respectively. While all of these administrative divisions have self-governing authority, they are classified differently depending on whether it is an urban or rural area as well as an urban area within a metropolitan area. Specifically, Si is an urbanized area, Gun is a rural area with smallsized towns, and Gu is an urbanized area within a metropolitan city in Korea.shows the spatial distribution of confirmed cases of COVID-19 at the county level as of 17 September 2020.
## Main outcomes
Our main outcome variable was the accrued number of positive cases starting from 20 January 2020 until 17 September 2020, collected per county, which is Si/Gun/Gu. There was no official county-level COVID-19 data, so we manually compiled the data from websites of each city, county, and district. The actual government websites from which we sourced the original data can be found in the(accessed on 18 September 2020). The total number of confirmed cases by 17 September 2020 was 22,657. Of this number, 1432 cases were people returning from overseas, so we excluded them from the analysis. Moreover, the website for Seoul Metropolitan City, Busan Metropolitan City, Incheon Metropolitan City, and North Jeolla Province had a separate category for patients who contracted the disease from other regions, such as other cities or from foreign countries. We excluded these cases, making the total number of confirmed cases 20,787.
## Explanatory variables
Of the variables that have the potential to affect the outcome variable, we chose indicators of density and connectivity as the explanatory variables. We calculated net density measures for the population and employment densities. Using the aggregated travel data with origination/destination(O/D) at the county level, social network theory was used to calculate the four different measures of connectivity. We further standardized the density and connectivity measures to enable robust comparison in the model results.
The urbanization-related indicators of density used in this research were population and employment densities. Data for these variables were drawn from the Korea Statistical Information Service (KOSIS). In order to accurately determine the effects of urban densities, we used the concept of net density, which refers to the number of housing units in a given area of land devoted to residential development. As discussed in the literature review, we took caution in calculating these density measures because the outskirts of urban
## Main outcomes
Our main outcome variable was the accrued number of positive cases starting from 20 January 2020 until 17 September 2020, collected per county, which is Si/Gun/Gu. There was no official county-level COVID-19 data, so we manually compiled the data from websites of each city, county, and district. The actual government websites from which we sourced the original data can be found in the(accessed on 18 September 2020). The total number of confirmed cases by 17 September 2020 was 22,657. Of this number, 1432 cases were people returning from overseas, so we excluded them from the analysis. Moreover, the website for Seoul Metropolitan City, Busan Metropolitan City, Incheon Metropolitan City, and North Jeolla Province had a separate category for patients who contracted the disease from other regions, such as other cities or from foreign countries. We excluded these cases, making the total number of confirmed cases 20,787.
## Explanatory variables
Of the variables that have the potential to affect the outcome variable, we chose indicators of density and connectivity as the explanatory variables. We calculated net density measures for the population and employment densities. Using the aggregated travel data with origination/destination(O/D) at the county level, social network theory was used to calculate the four different measures of connectivity. We further standardized the density and connectivity measures to enable robust comparison in the model results.
The urbanization-related indicators of density used in this research were population and employment densities. Data for these variables were drawn from the Korea Statistical Information Service (KOSIS). In order to accurately determine the effects of urban densities, we used the concept of net density, which refers to the number of housing units in a given area of land devoted to residential development. As discussed in the literature review, we took caution in calculating these density measures because the outskirts of urban areas in Korea are often dominated by forestry, agricultural lands, and other non-residential lands.
Failure to account for these inhabitable lands could result in the underestimation of urban density.
We used the 2020 travel O/D data at the county level on all transport modes provided by the Korea Transport Database (KTDB) to compute the connectivity measures. We used social network analysis (SNA) to calculate these measures. The form of the analysis is usually either an array of the social networks or a structure classified by different social network characteristics. A variety of social network analysis methods exist to quantify network structures, such as centrality, density, structural holes, concentration, and modularity.
In light of the COVID-19 pandemic, the analysis has been employed either to demonstrate the infection network of the confirmed patients and its structural characteristicsor to explore how key nodes can play an important role in social networks of COVID-19. These studies mainly relied on data from Twitter users and their followersto understand the spatial diffusion of the COVID-19 infectionor to visualize the COVID-19 pandemic risk.
Our approach for the social network analysis is different from the aforementioned studies. We applied the social network theory to estimate the spatial structure and its functional characteristics, resulting from the movement of people in daily life using transportation O/D data aggregated at the county level. A city is a place where people travel within and move to and from, and it should not be confined by its administrative boundary. Our goal was to identify the geospatial and relational characteristics which are embedded in the intricate and complex transportation networks and the movement of people influenced by their daily activities within the transportation networks.
We used the centrality measures to quantify how cities exist within a network. Specifically, we calculated the indicators of degree centrality, closeness centrality, betweenness centrality, and eigenvector centrality. The equations to calculate each of the centrality measures are shown below in.
## Indicators equation
## Degree
Centrality
[formula] g ∑ i=1 x ij , i = j ∑ g i=1 [/formula]
x ij : Number of connections node i has with (g − 1) nodes. The degree of centrality is defined as the summation of the number of direct links upon a node, and, using the number of different links a singular node has, the degree of the node's centrality can be quantified. In other words, degree centrality tells us if a node takes on a key role within the social network and is positioned in a place of importance. Cities with high-degree centrality can be interpreted as having high linkages, both direct and local, with other cities. Closeness centrality measures centrality based on the closeness of other nodes and differs from the degree of centrality in that it is calculated by adding the distances of the nodes that are both directly and indirectly linked. Since the distances between the nodes are considered, it is used to determine the global relationship of the social network as opposed to direct linkages. Betweenness centrality calculates where a node is situated in relation to other nodes, and if the value of centrality is high, the node can be understood as a mediating city between other cities. Finally, eigenvector centrality measures the comparative score of every node within a network based on the principle that connections to a node with high centrality must positively impact the centrality value of the node in question. In other words, a weight is placed on the connected node, which is then considered in calculating the node's centrality. Cities with high eigenvector centrality have a high probability of being linked to cities with a high degree of connectivity, which in turn, implies that these cities are more likely to come into contact with other contagious viruses. It is also important to note that if a node is linked to a city with a low connectivity value, the eigenvector-centrality value can decrease.
## Closeness centrality
[formula] 1 ∑ g i=1 d(N i ,N j ) ∑ g i=1 d N i , [/formula]
## Control variables
Our control variables included physical environment, socioeconomic factors, and other additional indicators. Hamidi et al.considered the vulnerable population, including the elderly, and the level of medical services and facilities, in addition to density, as the main factors affecting the spread of COVID-19 and mortality in the United States. Arbel et al.and Ehlertreported that socioeconomic factors other than density are closely related to the COVID-19 fatality rate. Identifying the effect of social distancing on trip reduction in U.S. metropolitan areas, Hamidi & Zandiatashbaremployed demographic attributes, park accessibility, as well as density. Teller, reviewing many empirical studies on the relationship between COVID-19 and urban density, reported that urban settlement environment, socio-economic factors, physical environment, and urban service level might be the main determinants of its spread. For example, access to parks can have more benefit than cost from the possible risk of COVID-19 disease outbreaks since they may play a role in mental health improvementand adaption of people to not-crowed outdoor activitiesduring COVID-19. Agnoletti et al.measured the unemployment rate as one of the control factors for the cases of COVID-19 by province.
We followed the methods used in previous studies to measure potential determinants of COVID-19 in Korea. Specifically, we included ratio variables for females and youth (20-29 years old) and the older adults (65+ years old) as demographic characteristics obtained from the KOSIS database. For neighborhood resources, measures of medical accessibility, nursing home service, and park accessibility were obtained from the public data portal (http://data.go.kr, accessed on 17 September 2020) and processed using ArcGIS version 10.6 (ESRI, San Diego, CA, USA), a geographic information system (GIS) software. Per capita growth regional domestic product (GRDP) was derived from the KOSIS database as a measure of neighborhood economic characteristics. Unlike the Italian study, we did not include the unemployment rate at the city-county-district level, as this measure was not available at the time of the study. Land area (km 2 ) was also included to control for differences in the spatial units. In addition, there was a sudden rise in infection associated with a super-spreader event at a particular church in the city of Daegu. As shown in, the confirmed cases from the Daegu metropolitan area accounted for the majority of accrued confirmed cases. Thus, we created a dummy variable to separately examine the effect of this particular region affected by the super-spreader event.
## Analytical approach
To estimate the numbers of COVID-19 cases over the study period, we developed a robust ordinary least squares (OLS) regression model and a robust negative binomial regression (NBR) model. The probability of being diagnosed with COVID-19 exhibits the same probability distribution of a coincidental event. In such a case, a Poisson regression model or a robust negative binomial model are more appropriate to use than a standard OLS model. When using the Poisson regression model, the underlying assumption is that the mean and the variance of an event occurring are identical to each other. Hence, if the actual analysis of the data shows over-dispersion, then there are obvious limitations to using this model. As such, when looking at the means and the variances of the two dependent variables used in this study, we found that the Poisson model was not suitable. Over-dispersion was detected in the means of the total number of confirmed cases and cases per 10,000 (91.17 and 3.55, respectively) as well as their variances .98 and 66.51, respectively). Also, not many cases of zero-patients were observed in our data; therefore, it was more suitable to use a robust negative binomial model than a zero-inflated Poisson model. Moreover, we chose to use the robust model to minimize the impact of heteroskedasticity.
## Analytical approach
To estimate the numbers of COVID-19 cases over the study period, we developed a robust ordinary least squares (OLS) regression model and a robust negative binomial re gression (NBR) model. The probability of being diagnosed with COVID-19 exhibits the same probability distribution of a coincidental event. In such a case, a Poisson regression model or a robust negative binomial model are more appropriate to use than a standard OLS model. When using the Poisson regression model, the underlying assumption i that the mean and the variance of an event occurring are identical to each other. Hence, i the actual analysis of the data shows over-dispersion, then there are obvious limitation to using this model. As such, when looking at the means and the variances of the two dependent variables used in this study, we found that the Poisson model was not suitable Over-dispersion was detected in the means of the total number of confirmed cases and cases per 10,000 (91.17 and 3.55, respectively) as well as their variances .98 and 66.51, respectively). Also, not many cases of zero-patients were observed in our data therefore, it was more suitable to use a robust negative binomial model than a zero-in flated Poisson model. Moreover, we chose to use the robust model to minimize the impact of heteroskedasticity.
Finally, this paper compared four models-two that differ by the type of model (OLS model and the NBR model) used and an additional two that differ in the definition used for the dependent variable (total number of positive cases and total number of positive cases per 10,000). Model A defined the dependent variable as the total number of con firmed positive cases, and the analysis was based on the OLS model. Model B, similar to Model A, used the same dependent variable but switched the analysis methodology to NBR. Model C, on the other hand, defined its dependent variable as the total number o positive cases per 10,000, and the OLS method was used for its analysis. Finally, we used the same dependent variable for Model D as we did for Model C but used the NBR method of analysis. All analyses were performed in Stata (version 16, StataCorp LLC, College Sta tion, TX, USA). Finally, this paper compared four models-two that differ by the type of model (OLS model and the NBR model) used and an additional two that differ in the definition used for the dependent variable (total number of positive cases and total number of positive cases per 10,000). Model A defined the dependent variable as the total number of confirmed positive cases, and the analysis was based on the OLS model. Model B, similar to Model A, used the same dependent variable but switched the analysis methodology to NBR. Model C, on the other hand, defined its dependent variable as the total number of positive cases per 10,000, and the OLS method was used for its analysis. Finally, we used the same dependent variable for Model D as we did for Model C but used the NBR method of analysis. All analyses were performed in Stata (version 16, StataCorp LLC, College Station, TX, USA).shows the summary statistics of the variables used in this research. The average number of confirmed COVID-19 at the city-county-district level was 91.17, and the confirmed cases per 10,000 population was 3.56. Their respective standard deviations were 196.05 and 8.16, which were more than twice the average. This indicates that there was larger between-group variability in the outcome measure. Also, the maximum values for cases and rate of COVID-19 were 1671 and 92.90, respectively, indicating non-normal distribution. The distribution of COVID-19 cases was skewed to the right; therefore, we standardized the density and connectivity indicators to allow direct comparison between the magnitudes of their impacts on the cases and their rates per 10,000 population for COVID-19.shows the results of the OLS and NBR models. Two different measures of outcome were estimated: one for the total number of confirmed cases (Models 1 and 2) and the other for the number of confirmed cases per 10,000 people . The indicator of dispersion, α was greater than zero, indicated that the data were over-dispersed. Hence, an NBR model appeared to be more suitable than a Poisson model.
# Results
## Descriptive summary
# Model results
In terms of the overall model fitness, the AIC and BIC values were smaller in the NBR models than the OLS models, indicating that the NBR models perform better than the OLS models. This implies that a random event could occur in the relationship, affecting the probability of infection. Hence, we focused our efforts on interpreting the NBR model results.
When it comes to the connectivity measures, degree centrality was most strongly correlated with the number of cases (coefficient = 45.62 for the OLS model and coefficient = 0.44 for the NBR model). However, none of the connectivity measures were statistically significant for the models that use infection rates. Of the density variables, only net population density was strongly associated with the number of cases (coefficient = 29.38 for the OLS model and coefficient = 0.31 for the NBR model). Again, none of the density measures were associated with the infection rates. Note: *** p < 0.001, ** p < 0.01, * p < 0.05, + p < 0.10.
Among the population characteristics, the number of cases was positively correlated with the proportion of females (coefficient = 21.16) but negatively associated with the proportion of older adults (coefficient = −0.052) in the NBR model. The OLS model as well as the models that used infection rates showed no statistically significant results with any of the population characteristics. All of the variables indicating neighborhood resources were strongly correlated with the number of cases in the NBR model (Model 2).
While the availability of a nursing home was positively correlated with the number of cases, the number of doctors per 1000 people and park area per 1000 people were both negatively correlated with the number of cases, indicating some protective effects of neighborhood resources against the COVID-19 outbreak. This finding is similar to the conclusion of Ma et al., who found that the benefits of visiting parks outweighed the risks associated with possible COVID-19 infection. Interestingly, the dummy variable indicating the city of Daegu was strongly correlated with both the number of cases and the infection rates across all models. This result indicates that the super-spreader event clustered around the Daegu area indeed was strongly associated with the spread of COVID-19 in the early phase of the outbreak.
# Discussion
Our results suggest that the total number of confirmed cases was more appropriate than the infection rates in explaining the relationship with urban spatial structure. One possible reason is that using infection rates in the model could have resulted in canceling out the effect of population density. Because our models included density measures that incorporated population counts in the denominator, dividing the number of confirmed cases by population counts could negate the effect of density and other urban-related factors, leading to the classic type II error. This indicates that further caution will be needed to account for possible biases when modeling the spread of infectious diseases.
Using the standardized coefficients, we found that the measures of connectivity were more positively related to COVID-19 infection than the measures of density (standardized coefficient = 0.31 for net population density; standardized coefficient = 0.44 for degree centrality). This result is somewhat consistent with findings from previous researchthat suggested that dense places do not necessarily lead to more infection but more connected places, as measured by a metropolitan size and an enplanement rate, were positively associated with the infection rates. Another study conducted in China also found that connectivity, as measured by the betweenness centrality of the road networks of 255 Chinese cities, was one of the key determinants of infection rates.
Interestingly, the eigenvector connectivity measure had a negative effect on the infection in our model. This was because the eigenvector centrality takes into account the level of connectivity of the connected node (city). Thus, even if the node is connected to a lower-networked city, it has a high degree of connectivity. Both closeness centrality and betweenness centrality were found to be statistically insignificant. This might be because the closeness-centrality measure captures the global connectivity of the social network; therefore, it might not fully capture all the variations in the network condition that are more related to the spread of the virus. Similarly, the betweenness centrality measures the degree to which a node is situated in between different cities within the social network. Again, we believe that this measure is irrelevant in explaining the network condition that is more vulnerable to virus dissemination. On the other hand, the degree of centrality which captures direct connection in the network was found to be the best predictor among other connectivity measures.
The net population density was positively and significantly associated with the virus infection. This is consistent with the results of previous studies that found a strong association between population density and infectious disease outbreak. Interestingly, net employment density was not statistically significant in the model. This is somewhat different from the findings of Ehlert, where both population density and employment density were significant associated with the spread of COVID-19. The reason for this difference could be because Ehlert's research was conducted in Germany and/or because he did not consider any indicators of connectivity in his research. In absence of the connectivity measure, employment density may actually capture the effect of the network connection between home and work. In addition, it is possible that our findings reflect the large-scale effect of social-distancing policies, where a transition to working from home gradually took place around August 2020. In fact, the Korean government never imposed strict lockdown policies that would have resulted in a complete closure of retail stores and businesses; however, our results suggest that activities and movements related to employment may have little impact on the spread of COVID-19. Further research should examine if there is a combined effect of social-distancing policies and employment density on COVID-19 outbreak.
# Limitations
There were a few limitations with the study. First, this study did not consider the influence of social-distancing policies implemented in Korea since February 2020. There were several versions of social-distancing policies throughout the study period, from school closure to restriction on social gatherings. We were unable to examine the effect of socialdistancing policies because our data were based on a cumulative record of COVID-19 cases, which did not include time information. The individual public health agency by province only releases aggregate data; therefore, we compiled the data manually from individual government websites at one point in time . Second, there is a slight possibility that the total number of cases in the early phase of the outbreak could be biased due to the availability of testing. However, some of the undercounting of confirmed cases were resolved as the government ramped up testing capacity early on. In fact, a number of international media praised Korea as one of the few countries with the most aggressive COVID-19 test program. Third, this study did not consider the impact of COVID-19 mortality because no reliable data were available at the appropriate spatial scale. If high-quality mortality data become available, it would be desirable to include mortality rates along with infection rates in future studies.
# Conclusions
This study provides evidence that connectivity seems to matter more than density itself. Contrary to the common belief, cities are not necessarily more vulnerable to the spread of the disease than the rural counterparts. Instead of focusing on restricting movements in all places, having stricter control over movements and activities in highly connected places could be more effective in dampening the spread of COVID-19.
We also found that direct and local degree centrality were more strongly associated with the spread of COVID-19 than the global closeness centrality. This suggests that future mitigation strategies may focus on restricting movements between cities that are directly linked. Despite having a lower effect size than the connectivity measure, the net population density was still statistically significant in explaining the dissemination of COVID-19. Therefore, we cannot ignore the fact that density, especially residential density, could still play a role in the spread of the disease.
Our study offers new insights into the current debate on the vulnerability of densely populated cities, especially from the perspective of an Asian country. If our results can be confirmed by prospective studies, they could have implications for developing mitigation strategies to contain future outbreaks. Connectivity and residential density may indeed reflect one of a constellation of factors that need to be taken into account for future planning to protect our cities against infectious disease threats in the foreseeable future. Our study provides important insights into how we might plan and build safer, healthier, and more resilient cities in a post pandemic world. |
Developments, application, and performance of artificial intelligence in dentistry – A systematic review
Background/purpose: Artificial intelligence (AI) has made deep inroads into dentistry in the last few years. The aim of this systematic review was to identify the development of AI applications that are widely employed in dentistry and evaluate their performance in terms of diagnosis, clinical decision-making, and predicting the prognosis of the treatment.
# Introduction
The exponential growth in science and technology has introduced different applications that are used daily, such as siri and alexa. These applications are found on the top of artificial intelligence (AI) and its components. The term AI is mostly associated with robotics. It describes how technology is used to develop a software or a machine that can easily mimic human intelligence and perform specific activities.
John McCarthy, a mathematician coined the term artificial intelligence in 1955, and widely recognized as the father of artificial intelligence. He chose this term to explain the potential of machines to perform tasks that can fall in the range of "intelligent" activities.In the year 1956, John McCarthy organized a famous Dartmouth conference, which was formally on the research project artificial intelligence, this is when this discipline originated. The conference sparked the crucial period, from 1950s to 1970s, wherein extensive research was done on AI.In 1978 Richard Bellman, an applied mathematician defined artificial intelligence as the automation of activities associated with human thinking abilities, which includes learning, decision making and problem solving.In the modern day world, artificial intelligence refers to any machine or technology that is able to mimic human cognitive skills like problem solving. To understand AI, it is important to know few of these key aspects .
- Artificial intelligence is termed as a capability of machines that exhibits a form of its own intelligence. The aim here was to develop machines that can learn through data so that they can solve the problems. - Machine learning is part of AI, which depends on algorithms to predict outcomes based on a dataset. The purpose of machine learning is to facilitate machines to learn from data so they can resolve issues without human input. - Neural networks are a set of algorithms that compute signals via artificial neurons. The purpose of neural networks is to create neural networks that function like the human brain. - Deep learning is a component of machine learning that utilizes the network with different computational layers in a deep neural network to analyze the input data. The purpose of deep learning is to construct a neural network that automatically identifies patterns to improve feature detection.Deep learning is also known as convolutional neural networks. They collect features from the abstracted layer of filters and are primarily used to process large and complex images. 5 Key aspects of artificial intelligence.
Application of AI in dentistry AI has been used mainly in dentistry to make the process of diagnosis more accurate and efficient, which is of atmost importance in achieving best results to the treatments provided along with superior quality patient care.
Dentists need to use all their acquired knowledge to diagnose and decide the best treatment option. They are also required to predict the prognosis where they need accurate clinical decision-making skills. However, in some cases, dentists do not have enough knowledge to make the right clinical decision in a limited period. AI applications can serve as their guide so that they can make better decisions and perform better.
Shortliffe, [bib_ref] Testing reality: the introduction of decisionsupport technologies for physicians, Shortliffe [/bib_ref] Chae et al., [bib_ref] The adoption of electronic medical records and decision support systems in Korea, Chae [/bib_ref] Schleyer et al., [bib_ref] Clinical computing in general dentistry, Schleyer [/bib_ref] reported that dentists have become dependent on computer applications to get insights for clinical decision making. The aim of this systematic review was to identify the development of AI applications that are widely employed in dentistry and to evaluate their performance in terms of diagnosis, clinical decision-making, and predicting the prognosis of the treatment.
# Materials and methods
## Data sources
This systematic review was carried after referring the guidelines for preferred reporting items for Systematic reviews and Meta-analyses extension for Diagnostic Test Accuracy (PRISMA-DTA). [bib_ref] Recommendations for reporting of systematic reviews and meta-analyses of diagnostic test accuracy:..., Mcgrath [/bib_ref] The literature for this paper was identified and selected by performing a thorough search in the electronic data bases like Pubmed, Medline, Embase, Cochrane, Google scholar, Scopus, Web of science, and Saudi digital library published over the past two decades (January 2000eMarch 15, 2020) by using keywords such as artificial intelligence in dentistry, deep learning, machine learning, artificial neural networks, convolutional neural networks, and computer-aided diagnosis. This search was based on the PICO (problem/patient/population, intervention/indicator, comparison, and outcome) elements [fig_ref] Table 1: Description of the PICO [/fig_ref].
## Resources selection
Full-length articles were retrieved. Hand searching and electronic searching was performed to go through the journals. The required data for this review was selected in two stages. In the first stage the articles were selected based on the title and abstracts related to our research topic. The preliminary search resulted in 1268 articles that were appropriate enough to address the paper's aim. Due to duplication, 182 articles were removed. Hence, we retrieved 1086 articles for the second stage of selection. Next, the following criterion was applied.
Criteria for considering studies for this review Inclusion criteria 1. The article must be focused on AI and its application should be related to dentistry. 2. There must be some predictive or measurable outcomes so they can be quantified. 3. There has to be a proper mention of datasets that are used to assess a model.
## Exclusion criteria
1. The articles that are related to non-AI areas.
2. Uistaploaded articles that were unpublished. 3. Articles that consisted of only abstracts without the full text. 4. Articles that were not written in English.
This criteria cut down the number of articles to 46. The name of the journals and names of the authors were hidden and spread among the authors. A critical assessment was carried out for all the articles by following QUADAS-2 (Quality Assessment and Diagnostic Accuracy Tool) guidelines, a tool for quality assessment of the studies on diagnostic accuracy. [bib_ref] QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies, Whiting [/bib_ref] 3 more articles were excluded due to disagreement from the authors. Eventually, this systematic review performed qualitative synthesis on 43 articles [fig_ref] Figure 2: Flow chart for screening and selection of articles [/fig_ref].
All the articles were read fully. The years of these articles were taken in account to study the progress of AI trends that were developed and evolved over the years in dentistry.
# Results
This systematic review included 43 research articles that were analyzed for quantitative data. The analysis of the literature revealed that most of the studies were conducted in the past 12 years [fig_ref] Figure 3: Trends of research on artificial intelligence in dentistry [/fig_ref]. The trends showed that there is a gradual increase in the research related to artificial intelligence in dentistry.
The studies that were included in this systematic review were mainly on the application of AI for detection and diagnosis of dental caries, diagnosis of proximal dental caries, tooth detection and numbering, detection of vertical root fractures, detection of apical lesions, locating minor apical foramen, assessment of root morphology, diagnosis of salivary gland diseases, diagnosis of maxillary sinusitis, maxillofacial cysts, cervical lymph nodes metastasis, detection of osteoporosis, predicting diagnosis of orthodontic extractions, assessing need for orthodontic treatments, determining the growth and development of cervical vertebrae stages, cephalometric analysis, identifying cephalometric landmarks, diagnosis of orthognathic surgeries, assessing the impact of orthodontic treatment on the facial attractiveness, gender determination using mandibular morphometric parameters, estimating the age based on third molar development, classifying the cancerous tissues, predicting postoperative facial swelling following extractions, detection of periodontal bone loss, detecting the degree of alveolar bone loss.
Most of the studies have used convolutional neural networks (CNNs) and artificial neural networks (ANNs). Some of the studies were based on bayesian network (BN) and probabilistic neural networks (PNNS). The purpose of these neural networks is to determine computer tomography (CT) images, cone beam computed tomography (CBCT) images, lateral cephalometric radiographs, bitewing radiographs, facial photos, panoramic radiographs (OPG). [fig_ref] Table 2: Details of the studies that have used AI based models in various... [/fig_ref].
The studies that have been analyzed in this systematic review revealed that the AI technology has been widely used in different specialties of dentistry. AI technology has been applied in 18 studies on oral & maxillofacial radiology, [bib_ref] Osteoporosis detection in panoramic radiographs using a deep convolutional neural network-based computer-assisted..., Lee [/bib_ref] [bib_ref] An effective teeth recognition method using label tree with cascade network structure, Zhang [/bib_ref] [bib_ref] Detection and diagnosis of dental caries using a deep learning-based convolutional neural..., Lee [/bib_ref] [bib_ref] Caries detection with near-infrared transillumination using deep learning, Casalegno [/bib_ref] [bib_ref] Preliminary study on the application of deep learning system to diagnosis of..., Kise [/bib_ref] ,36e41,44e48,51 11 studies on orthodontics and dentofacial orthopedics, [bib_ref] Artificial neural network modeling for deciding if extractions are necessary prior to..., Xie [/bib_ref]
## Risk of bias assessment
All the studies were based on assessing the diagnostic accuracy of AI in various fields of dentistry. Hence, risk of bias assessment was assessed using QUADAS-2 assessment tool, which is frequently used tool in the literature (Supplementary [fig_ref] Table 1: Description of the PICO [/fig_ref]. The studies conducted on human beings for determining the reference standard were rated as high risk. In the present analysis, 55.81% of studies reported high risk of bias for the reference standard. Since the data feeding in AI technology was highly standardized, AI had no effect on the flow, and time frame in the final output, hence categorized as low risk categories. In the present systematic review, low risk of bias was reported in index test (81%) and (74.41%) in flow and timings. Some studies conducted on the cadaveric samples and extracted specimens, under the applicability domain were considered as high risk for patient selection and index test domain (Supplementary [fig_ref] Table 2: Details of the studies that have used AI based models in various... [/fig_ref]. Even though, comparable results were obtained for the applicability arm of the QUADAS-2 (Figs. 4 and 5).
# Discussion
AI is modernizing the traditional aspects of dentistry. AI based systems are often used for designing automated software programs that streamlines the diagnosis and data management in dentistry. [bib_ref] Clinical computing in general dentistry, Schleyer [/bib_ref] Mostly they are clinical decision support systems that assist and guide experts to make better decisions. These systems have been used for better diagnosis, treatment planning and also for predicting the prognosis. [bib_ref] Clinical decision support systems: perspectives in dentistry, Mendonça [/bib_ref] The demand for these systems is booming due to their effectiveness in providing explanations and reasoning. [bib_ref] A casebased decision support system for individual stress diagnosis using fuzzy similarity..., Begum [/bib_ref] AI has revolutionized in the field of dentistry and making the dentist's task easier. The clinical decision support systems that work on the AI technology are mainly designed to provide expert support to the health professionals. [bib_ref] Artificial intelligence: contemporary applications and future compass, Khanna [/bib_ref] Clinical decision support systems is defined as, any computer program that has been designed to help health professionals in making clinical decisions, and also deals with the medical data or with the knowledge of medicine necessary for interpreting such data. [bib_ref] Testing reality: the introduction of decisionsupport technologies for physicians, Shortliffe [/bib_ref] In this systematic review we analyzed studies on the application of AI technology in dentistry and evaluated their performance in terms of diagnosis, clinical decisionmaking, and predicting the prognosis of the treatment.
## Application of ai technologies in the specialty of oral and maxillofacial radiology and diagnostics
Zhank et al., [bib_ref] An effective teeth recognition method using label tree with cascade network structure, Zhang [/bib_ref] reported the use of AI-based CNNs and evaluated effective teeth recognition by relying on the label tree along with cascade network structure. The model demonstrated a high precision of 95.8%. A study by Tuzoff et al., [bib_ref] Tooth detection and numbering in panoramic radiographs using convolutional neural networks, Tuzoff [/bib_ref] showed very similar results when utilized an AIpowered CNNs model for teeth identification, which was then arranged numerically. This computer-aided diagnostic technique displayed a mean sensitivity of 0.987 and precision 0.9945. This output was similar to that of an expert.
Chen et al., [bib_ref] A deep learning approach to automatic teeth detection and numbering based on..., Chen [/bib_ref] applied the CNNs to detect the teeth number in intra oral periapical films and then to identify the tooth. The model demonstrated very high precision. The results indicated that AI technologies make it convenient for clinicians to do their job. They do not have to enter the details manually. Using these automated systems dentists can enter their dental charts digitally, resulting in higher efficiency.
AI technology in the detection of dental caries has demonstrated excellent results which were reflected in the study done by Lee et al., [bib_ref] Detection and diagnosis of dental caries using a deep learning-based convolutional neural..., Lee [/bib_ref] who reported applying of CNN algorithms for detection and diagnosis of dental caries on periapical radiographs. The result of the application demonstrated considerably good performance. Similar results were seen in the study done by Casalegno et al., [bib_ref] Caries detection with near-infrared transillumination using deep learning, Casalegno [/bib_ref] who used deep learning model designed for the detection and Artificial intelligence in dentistry localization of dental lesions in near-infrared transillumination (TI) images which revealed promising results. Schwendicke et al., [bib_ref] Deep learning for caries lesion detection in nearinfrared light transillumination images: a..., Schwendicke [/bib_ref] reported using of near-infrared-light transillumination (NILT) images for diagnosing dental caries and showed that the performance of this AI based models was satisfactory. Devito et al.,used an AI based ANN model for diagnosing proximal caries using the bitewing radiographs and found a quite encouraging results. Study by Hung et al., [bib_ref] Application of machine learning for diagnostic prediction of root caries, Hung [/bib_ref] reported AI technology for predicting root caries and demonstrated excellent results. Ekert et al., [bib_ref] Deep learning for the radiographic detection of apical lesions, Ekert [/bib_ref] was successful in detecting apical lesion when they applied CNNs to detect apical lesions (ALs) on panoramic dental radiographs.
## Artificial intelligence in dentistry
Hiraiwa et al., [bib_ref] Preliminary study on the application of deep learning system to diagnosis of..., Kise [/bib_ref] reported of applying CNNs for detection of sjögren's syndrome (SjS) on CT images and compared the results with the performance of radiologists and showed a higher diagnostic performance. In another study by Murata et al., [bib_ref] Deep-learning classification using convolutional neural network for evaluation of maxillary sinusitis on..., Murata [/bib_ref] the authors applied the deep learning system for diagnosing maxillary sinusitis on panoramic radiography. The diagnostic performance of this system was sufficiently high. These results were similar to the study conducted by kim et al., [bib_ref] Deep Learning in diagnosis of maxillary sinusitis using conventional radiography, Kim [/bib_ref] when compared the performance with experienced radiologists.
Ariji et al., [bib_ref] CT evaluation of extranodal extension of cervical lymph node metastases in patients..., Ariji [/bib_ref] applied CNN system for deep learning image classification for diagnosing lymph node metastasis on (CT) Images and showed higher diagnosing accuracy, sensitivity, and specificity. These results were similar to the study conducted by Hung et al., [bib_ref] Application of machine learning for diagnostic prediction of root caries, Hung [/bib_ref] who assessed the performance of deep learning classification in diagnosing extra nodal extension of cervical lymph node metastases in CT images. Both the studies showed similar performance or even higher performance when compared to professional radiologists.
Lee et al., [bib_ref] Osteoporosis detection in panoramic radiographs using a deep convolutional neural network-based computer-assisted..., Lee [/bib_ref] evaluated the efficiency and performance of AI in diagnosis and detection of osteoporosis. In this study deep convolutional neural network (DCNN) based computer-assisted diagnosis (CAD) systems was applied for detection of osteoporosis, using panoramic radiographs and exhibited very promising results. This was well above in par with experienced oral and maxillofacial radiologists in detecting osteoporosis. These results were similar to the results done by Lee et al., [bib_ref] Evaluation of transfer learning with deep convolutional neural networks for screening osteoporosis..., Lee [/bib_ref] who used deep convolutional neural networks (DCNN) for detecting osteoporosis in dental panoramic radiographs.
Application of AI technologies in the specialty of orthodontics and dentofacial orthopedics Accurate diagnosis, treatment planning and prediction of prognosis are the key factors for successful orthodontic treatment. AI technology has been applied for deciding if extractions are necessary prior to the orthodontic treatment. In a study by Xie et al., [bib_ref] Artificial neural network modeling for deciding if extractions are necessary prior to..., Xie [/bib_ref] artificial neural network (ANN) model was applied for deciding if extractions are necessary using lateral cephalometric radiographs. The results were quite promising. Jung et al., [bib_ref] New approach for the diagnosis of extractions with neural network machine learning, Jung [/bib_ref] showed 92% accuracy using AI expert system for deciding on permanent tooth extraction, using lateral cephalometric radiographs. The results of both the studies are suggestive that the AI modes were effective and accurate in predicting the need for extraction. These models can be used as a tool for making decisions in clinical practice. High accuracy was seen in the study by Thanathornwong, 24 who suggested AI model based on bayesian network (BN) for assessing the need for orthodontic treatment.
Various studies have been conducted to demonstrate AI technologies and its application in identifying cephalometric landmarks. Park et al., [bib_ref] Automated identification of cephalometric landmarks: part 1-comparisons between the latest deep-learning methods..., Park [/bib_ref] compared the efficiency and accuracy of the updated deep-learning algorithms for automatic identification of cephalometric landmarks using cephalometric radiographs. The results revealed that the system was extremely accurate in the computation of the landmarks. Studies conducted by Kunz et al., [bib_ref] Artificial intelligence in orthodontics: evaluation of a fully automated cephalometric analysis using..., Kunz [/bib_ref] and Hwang et al., [bib_ref] Automated identification of cephalometric landmarks: Part 2-Might it be better than human?, Hwang [/bib_ref] showed excellent accuracy in identifying the landmarks similar to the trained human examiners using a specialized artificial intelligence (AI) algorithm and deep learning based automated identification system respectively. Yu et al. [bib_ref] Automated skeletal classification with lateral cephalometry based on artificial intelligence, Yu [/bib_ref] demonstrated excellent results with automated skeletal classification with lateral cephalometry based on the AI Model. The results of the above mentioned studies indicates that, these systems prove to be a viable option for repeatedly identifying multiple cephalometric landmarks.
Establishing of accurate diagnosis and treatment planning in orthognathic surgery is the most important step for the success of the treatment.Arnet et al., [bib_ref] Facial keys to orthodontic diagnosis and treatment planning. Part I, Arnett [/bib_ref] in his literature on facial keys to orthodontic diagnosis and treatment planning suggested that if diagnosis is incorrect, the patient esthetics may further deteriorate creating a major problem. This suggests that diagnosis is an important aspect for the dentist to analyze the problems of the patient accurately. AI technology is striving to make dentists job much accurate and precise. Choi et al., [bib_ref] Artificial intelligent model with neural network machine learning for the diagnosis of..., Choi [/bib_ref] reported the use of new artificial intelligence model to decide the case for surgery/ non-surgery using the lateral cephalometric radiographs. He showed that the system was very effective with 96% success rate in diagnosing the surgery/non-surgery cases. This model has shown promising results, hence can be applied for the diagnosis of orthognathic surgery cases.
Hagg et al., [bib_ref] Maturation indicators and the pubertal growth spurt, Hägg [/bib_ref] stated that determining the chronological age of the patient is not alone sufficient for estimating the actual growth time, hence various skeletal maturation indicators have been developed for this task. Determination of the growth and development, and estimation of the skeletal maturation stages have been used to predict the time of pubertal development, determining the growth rate and for mainly estimating the remaining growth and development potential of an individual as mentioned by Flores-mir et al. [bib_ref] Use of skeletal maturation based on hand-wrist radiographic analysis as a predictor..., Flores-Mir [/bib_ref] These are usually determined by using hand-wrist radiographs, cephalometric analysis, and with the help of maturation stages of cervical vertebra. AI technology has also been applied for determining the growth and development by cervical vertebrae stages. Kok H et al., [bib_ref] Usage and comparison of artificial intelligence algorithms for determination of growth and..., Kök [/bib_ref] showed a mean accuracy of 77.02%, using artificial intelligence algorithms for determining the growth and development by cervical vertebrae stages when applied on the cephalometric radiographs.
## Application of ai technologies in the specialty of endodontics
The success of root canal treatment mainly depends on accuracy of working length determination. The prognosis of the treatment can only be ensured when instrumentation terminates at the apical constriction. [bib_ref] The role of apical instrumentation in root canal treatment: a review of..., Baugh [/bib_ref] Saghiri et al. [bib_ref] A new approach for locating the minor apical foramen using an artificial..., Saghiri [/bib_ref] used artificial neural network (ANN) system in determining the working length and showed exceptional accuracy of 96% which is higher than the accuracy compared to professional endodontists. These results were similar to the study by Saghiri et al., [bib_ref] The reliability of artificial neural network in locating minor apical foramen: a..., Saghiri [/bib_ref] where they used the ANN system for locating the minor apical foramen, with an accuracy of 93%.
In endodontics, AI is used to diagnose vertical root fractures. A study employed by Johari et al., [bib_ref] Detection of vertical root fractures in intact and endodontically treated premolar teeth..., Johari [/bib_ref] who used probabilistic neural network (PNN) for the diagnosis of vertical root fractures. This PNN system displayed excellent performance with an accuracy of 96.6%. Similarly convolutional neural network in detecting vertical root fracture, showing a highly encouraging precision was conducted by Fukuda et al. [bib_ref] Evaluation of an artificial intelligence system for detecting vertical root fracture on..., Fukuda [/bib_ref] These assessments indicate that AI-based models are incredibly effective when it comes to the detection of vertical root fractures on CBCT images and panoramic radiographs.
## Application of ai technologies in the specialty of periodontics
Periodontal diseases are one of the most common oral diseases affecting the mankind. It is a known fact that this is one of the main reasons for the early loss of teeth. It is well reported by Lee et al., [bib_ref] National dental policies and socio-demographic factors affecting changes in the incidence of..., Lee [/bib_ref] that continuous progression of the disease will eventually lead to the loss of teeth in the adults. Various studies have been done to ascertain AI technology application to diagnose and predict periodontal diseases. Lee et al., [bib_ref] Diagnosis and prediction of periodontally compromised teeth using a deep learning-based convolutional..., Lee [/bib_ref] reported use of CAD system, based on a deep convolutional neural network (CNN) algorithm for diagnosing and predicting the teeth that are compromised with periodontal health. The outcome were quite acceptable with a mean predictive accuracy of 78.9%. Yauney et al., [bib_ref] Automated process incorporating machine learning segmentation and correlation of oral diseases with..., Yauney [/bib_ref] used an AI based system based on CNNs for correlating poor periodontal health with systemic health outcomes and reported that, AI can be used for automated diagnoses and can also be useful for screenings for other diseases. Krois et al.,used CNNs to detect periodontal bone loss (PBL) on panoramic dental radiographs. The results of this study were similar to that of the expert opinions. This system can still help in reducing the dentist's diagnostic efforts.
## Application of ai technologies in the specialty of oral and maxillofacial surgery
It is estimated that every year there are around 657,000 new cases detected of cancers of the oral cavity and pharynx and is also a reason for 330,000 deaths as noted in the article of oral cancer by WHO. 63 AI technology has been used for detecting cancers. There is revolutionary development and refinement of convolutional neural networks that have demonstrated improved ability for automated cancer detection as seen in the study by Xu et al.Aubreville et al., [bib_ref] Automatic Classification of cancerous tissue in laser endomicroscopy images of the oral..., Aubreville [/bib_ref] showed extremely positive and promising results when employed CNNs for an automatic approach for diagnosing Oral Squamous Cell Carcinoma when used with confocal laser endomicroscopy images. The study indicated that AI model will be helpful for early diagnosis. AI technology has also been used for predicting postoperative facial swelling after extraction of teeth. Zhang et al., [bib_ref] Predicting postoperative facial swelling following impacted mandibular third molars extraction by using..., Zhang [/bib_ref] used an artificial intelligence model based on ANN for predicting the postoperative facial swelling following the extraction of impacted mandibular 3rd molars. The model demonstrated excellent results and will be of great importance for clinicians for predicting the prognosis of the treatment.
## Application of ai technologies in the forensic odontology
Forensic odontology is relatively new, but it has made a stellar contribution to the field of dentistry. A dentist plays an important role when they have to identify people for child abuse, crime, sexual assault, mass calamities, and other legal issues. Their moral duty compels them to provide justice to the victims and their families, especially when there is no other evidence other than the dental remains. AI technology has been applied in this field and has shown excellent results.
De Tobel et al., [bib_ref] An automated technique to stage lower third molar development on panoramic radiographs..., Tobel [/bib_ref] used automated technique based on CNNs for staging lower third molar development for estimating the age of a person after applying on panoramic radiographs. The system showed remarkable results, when compared to the trained examiners. Patil et al., [bib_ref] Artificial neural network for gender determination using mandibular morphometric parameters: a comparative..., Patil [/bib_ref] used ANNs to determine gender using panoramic radiographs, the results were quite promising. This system is very useful as it automates and eases the method of identifying unknown gender or age with minimal errors. Niño-Sandoval et al., [bib_ref] Use of automated learning techniques for predicting mandibular morphology in skeletal class..., Niño-Sandoval [/bib_ref] reported an AI model based on ANNs for predicting the mandibular morphology and demonstrated promising results. Hence AI can be used effectively in forensic dentistry.
Uses of AI in dentistry based on the conclusions from the articles reviewed in the paper - AI systems can assist the clinicians so they can offer high-quality dental care to their patients. - Dentists can use AI systems as an ancillary tool for increasing the accuracy of diagnosis, treatment planning, and predicting the treatment outcomes. - Non-specialty dentists can receive diagnostic support via the deep-learning systems. - Automated systems can save a lot of time and increase the efficiency of the clinicians (for e.g. automatic completion of electronic dental records by identifying the tooth and numbering). - The use of these systems for secondary opinions can improve the accuracy of diagnosis. - These systems provide a great deal of value for forensic diagnosis.
# Conclusion
AI has revolutionized dentistry in the last few years. Studies show that these AI-powered automated systems performed extremely well in various scenarios. Few authors found them to be more accurate than even dental specialists. Although these outcomes do not make them better than the dentists, they do establish that AI can be considered for clinical applications. These systems bring terrific value to the table by improving the accuracy of diagnosis, enhancing clinical decision-making, and predicting the treatment prognosis which can help the clinicians in rendering best quality care to their patients. There are also documented studies that have also reported that these automated systems are of greater value for screening the patients for osteoporosis, oral cancer, and metastasis of the lymph nodes. This is a priceless benefit because it can help professionals to diagnose cases in the early stages, which in turn can save many lives. Although AI is widely used in various fields of dentistry, some specialties such as pedodontics and oral pathology still lack the development and application of AI technology.
## Conflicts of interest
The authors have no conflicts of interest relevant to this article.
[fig] Figure 2: Flow chart for screening and selection of articles. [/fig]
[fig] Figure 3: Trends of research on artificial intelligence in dentistry. [/fig]
[fig] Figure 4: Assessment of individual risk of bias domains. [/fig]
[fig] Figure 5: Concerns regarding applicability. [/fig]
[table] Table 1: Description of the PICO (PZ Population, IZ Intervention, CZ Comparison, OZ Outcome) elements. [/table]
[table] Table 2: Details of the studies that have used AI based models in various specialties of dentistry for diagnosis, treatment planning, clinical decision making, predicting the need for treatment, and predicting the prognosis. [/table]
|
Development of a sublingual allergy vaccine for grass pollinosis
Grass pollen is a very common cause of allergic rhinitis and asthma. The only treatment targeting the underlying causes of allergy is immunotherapy (IT). Sublingual immunotherapy (SLIT) has been introduced to solve the problem of systemic reactions to subcutaneous IT (SCIT). This article evaluates the characteristics of the allergen extract, Staloral, in terms of practical administration, effectiveness, safety, and mechanism of action. Efficacy data were obtained from double-blind, placebo-controlled studies using Staloral in patients sensitized to grass pollen, while practical administration, cost-effectiveness, and mechanism of action data were provided by well designed studies. The efficacy and safety of Staloral, as demonstrated by review of published studies which used doses up to 1125 times those administered with SCIT, shows that this allergen extract has optimal characteristics for treating patients with seasonal allergies due to grass pollens. The main mechanism of action is the interaction between dendritic cells of the oral mucosa and the subsequent tolerance induced in T-cells.
# Introduction
Grass pollen is a major source of respiratory allergy throughout the world. [bib_ref] Allergenic pollen and pollen allergy in Europe, D'amato [/bib_ref] [bib_ref] The role of pollens in allergy, Freye [/bib_ref] Grasses include a high number of species, which have variable importance in different geographical areas. In Europe, Dactylis glomerata, Poa pratensis, Lolium perenne, and Anthoxantum odoratum are homogeneously distributed, while Phleum pratense is the dominant grass in Northern regions and in the UK, and Cynodon dactylon occurs only in Southern regions [fig_ref] Table 1: Distribution of grass species in europe [/fig_ref]. In the US, the species cited above are diffusely spread, while C. dactylon is present only in the Northeast, Midwestern, and South/Southeast regions.
Allergy to grass pollen can be treated by symptomatic drugs, namely antihistamines, topical and systemic corticosteroids, and antileukotrienes, but only specific immunotherapy (IT) is able to act on the cause of allergy and to modify the natural history of the disease. [bib_ref] World Health Organization Position Paper. Allergen immunotherapy: Therapeutic vaccines for allergic diseases, Bousquet [/bib_ref] IT was introduced in 1910, 4 demonstrated to be effective by a controlled trial in 1954, [bib_ref] Prophylaxis of summer hay-fever and asthma: A controlled trial comparing crude grass-pollen..., Frankland [/bib_ref] and established further as effective in meta-analyses of studies of traditional, subcutaneous immunotherapy (SCIT), [bib_ref] Allergen immunotherapy for asthma, Abramson [/bib_ref] [bib_ref] Allergen injection immunotherapy for seasonal allergic rhinitis, Calderon [/bib_ref] as well as sublingual immunotherapy (SLIT). [bib_ref] Sublingual immunotherapy for allergic rhinitis: Systematic review and meta-analysis, Wilson [/bib_ref] Most immunotherapy studies were performed using extracts containing multiple grasses, but there are trials using only D. glomerata, P. pratense, or L. perenne, on the basis that they have all the relevant grass allergens. However, thus far, there are no immunologic data demonstrating that extracts of single grasses are equivalent to multiple grass extracts in terms of IgE-binding capacity in sera from grass allergic patients. This suggests that multiple grass extracts may fit better with individual patients' patterns of sensitization. 9 submit your manuscript | www.dovepress.com
## Dovepress
## Dovepress
## 100
Frati et al Allergen extracts to be used for IT should have all the relevant allergens from the naturally occurring source, in a known and reproducible composition. For many years extracts used for SCIT were measured in arbitrary units or in protein nitrogen units, but in the 1990s SLIT was introduced as a consequence of allergen extract standardization, that is, with a known composition, and constantly reproducible on a lot-to-lot basis, thereby ensuring adequate quality of allergen products. [bib_ref] Allergen standardization, Platts Mills [/bib_ref] Availability of highquality, standardized extracts meets the need for indications of IT, ie, moderate to severe rhinitis and mild to moderate asthma. [bib_ref] World Health Organization Position Paper. Allergen immunotherapy: Therapeutic vaccines for allergic diseases, Bousquet [/bib_ref] This paper describes the development of a standardized grass pollen extract ( Staloral, Stallergenes, Antony, France) for the treatment of grass pollen-induced rhinitis and asthma.
## Standardization of staloral
The process of standardization is summarized in [fig_ref] Figure 1: microscopic identification of components from raw materials Comparison of the protein profile... [/fig_ref]. The index of reactivity (IR) is the biologic unit used; a concentration of 100 IR is determined by the capacity of the allergen to elicit by skin prick test a geometric mean wheal size of 7 mm diameter in 30 patients sensitized to the corresponding allergen. During the standardization procedure, the material is in the liquid phase. The product must be stored between 2°C and 8°C. At room temperature, the extract is stable for three months.
Several allergen extracts for SLIT are currently marketed in Europe, and vary in the amount of allergens contained. Almost all of the commercially available SLIT vaccines in Europe are standardized biologically or immunologically, 11 but because the standardization methods are based upon in-house references, the different extracts are labeled in units that differ from manufacturer to manufacturer. On the other hand, during the last few years, labelling of the content of major allergens in µg has been a significant improvement, enabling a rough comparison of different therapeutic regimens.
## Practical administration
SLIT allergen extract solution can be delivered by dropcounters, pre-dosed actuators (mini-pumps), or disposable single-dose vials. Staloral100 IR was initially administered by a drop-counter, but subsequently a mini-pump pre-dosed actuator was introduced, and is the same as the one used for Staloral 300 IR. The allergen extract is usually taken in the morning before breakfast, is kept under the tongue for few minutes, and then swallowed. The acronym SLIT, unless otherwise stated, indicates the sublingual-swallow modality.
## Administration regimens
SLIT traditionally involves an uptitration phase followed by maintenance treatment with the maximal dose. [bib_ref] Sublingual immunotherapy for allergic rhinitis: Systematic review and meta-analysis, Wilson [/bib_ref] In the past the uptitration phase has lasted 4-6 weeks, but this has progressively been shortened. The extract is available in vials at concentrations of 10, 100, and 300 IR. Other concentrations are available for sensitized patients. The patient must start with the lowest concentration and gradually increase, according to the physician's prescription, until the maintenance dose is reached. Rush and ultra-rush inductions have also been proposed for routine use, based on the favorable safety profile of Staloral, including in children under the age of five years. [bib_ref] Safety of SLIT in children aged 3 to 7 years, Fiocchi [/bib_ref] Ultra-rush schemes with a build-up shorter than two hours have been reported in adults [bib_ref] A pilot study of feasibility of ultra-rush (20-25 minutes) sublingual-swallow immunotherapy in..., Rossi [/bib_ref] and children, [bib_ref] Safety and tolerability of ultra-rush induction, less than one hour, of sublingual..., Tripodi [/bib_ref] with satisfactory results.
The maintenance dose is generally the same for all patients, based on clinical results. The maintenance dose recommended for Staloral is eight pressures corresponding to 16.8 µg of the major allergen Phl p 5 for six months/year.
The use of the same dose for all patients is justified on the basis that, in contrast with SCIT, the administration of very high amounts of allergens does not provoke severe side effects. The maintenance dose could be administered at variable time intervals but the once-daily simplifies the regimen by assuming the dose as done with almost all drugs.
## Criteria for starting and stopping treatment
Staloral can be administered either pre-coseasonally ( starting 2-4 months before and stopping at the end of the pollen season), coseasonally (only during the pollen season), or continuously. [bib_ref] Administration regimens for sublingual immunotherapy to pollen allergens: What do we know?, Lombardi [/bib_ref] In the pre-co-seasonal and co-seasonal regimen, no dose reduction during the pollen season is usually necessary, since there is no decrease in safety. [bib_ref] Safety of sublingual immunotherapy started during the pollen season, Ariano [/bib_ref] A recent study reported that a coseasonal schedule of 300 IR/day for three years had a carryover effect following Staloral discontinuation. 17
## Clinical efficacy
The first studies on SLIT used very low or low doses, [bib_ref] Noninjection routes for immunotherapy, Canonica [/bib_ref] but it subsequently became apparent that 50-to 100-fold higher doses than those administered for SCIT were needed to ensure clinical efficacy, [bib_ref] Allergic Rhinits and its Impact on Asthma (ARIA), Bousquet [/bib_ref] thereby introducing the concept of "high-dose SLIT". Staloral can be classified in this category because the ratio with the extract used for SCIT is higher than 50 (in most studies higher than 300) for any allergen used. Thus far, the results of eight studies of SLIT with grass pollen are available, all demonstrating significantly lower mean symptom/ medication [bib_ref] A double-blind, placebo-controlled trial by the sublingual route of immunotherapy with a..., Sabbah [/bib_ref] [bib_ref] Clinical efficacy of sublingual-swallow immunotherapy: A double-blind, placebo controlled trial of a..., Clavel [/bib_ref] [bib_ref] Sublingual-swallow immunotherapy (SLIT) with a standardized five-grass-pollen extract (drops and sublingual tablets)..., Pradalier [/bib_ref] [bib_ref] Randomized controlled trial of high-dose sublingual immunotherapy to treat seasonal allergic rhinitis, Smith [/bib_ref] [bib_ref] Sublingual immunotherapy in pollen-induced seasonal rhinitis and conjunctivitis: A randomized controlled trial, Mosges [/bib_ref] and children [bib_ref] Efficacy of grass pollen sublingual immunotherapy for three consecutive seasons and after..., Ott [/bib_ref] [bib_ref] Clinical efficacy of sublingual-swallow immunotherapy: A double-blind, placebo controlled trial of a..., Clavel [/bib_ref] [bib_ref] Sublingual immunotherapy and influence on urinary leukotrienes in seasonal pediatric allergy, Yuksel [/bib_ref] [bib_ref] Efficacy and safety of high-doses sublingual immunotherapy in ultra-rush scheme in children..., Stelmach [/bib_ref] treated with Staloral compared with placebo-treated controls. [fig_ref] Table 2: Published studies of SLiT with Staloral grass pollen [/fig_ref] reports the main data from the published studies evaluating Staloral for grass pollen allergy. In one additional, randomized, open-label study, different doses of Staloral were compared, demonstrating that the 300 IR maintenance dose had optimal effectiveness in terms of immunologic response. 27
## Cost-effectiveness
Currently, cost-effectiveness is an important issue in health care. Some studies have addressed this question in SLIT treatment with Staloral. The first published study evaluated the cost-effectiveness of SLIT in 135 children with allergic rhinitis and asthma (89 of whom were sensitized to grass pollen) by assessing direct costs (drugs, specialists visits, and SLIT) and indirect costs (costs resulting from absences from school and work). A substantial reduction was found in all outcome measures after the start of SLIT treatment, indicating that high-dose SLIT may be effective in reducing the economic burden of allergic rhinitis and asthma. [bib_ref] Cost effectiveness of sublingual immunotherapy in children with allergic rhinitis and asthma, Berto [/bib_ref] A second study was done using a decision tree populated with epidemiologic and resource utilization data for approximately 2200 adults with pollen allergy attending 25 Italian centers over a six-year period. [bib_ref] Economic evaluation of sublingual immunotherapy vs symptomatic treatment in adults with pollen-induced..., Berto [/bib_ref] Assessment criteria for the pharmacoeconomic analysis included direct medical costs from the National Health Service perspective (ie, visits, diagnostic procedures, drugs, SLIT, and hospitalizations), and indirect costs (lost working days and patient out-of-pocket expenses) from the societal perspective. Endpoints used for effectiveness included number of patients improved by treatment and number of asthma cases avoided. Finally incremental costs per improved patient and incremental costs per asthma case avoided were measured. A mean cost per patient treated over a period of six years was calculated for each therapeutic strategy and for each of the two perspectives studied. The SLIT strategy resulted in less expense in term of both direct and indirect costs. The break-even point of SLIT, ie, the time at which the overall cost of treatment for SLIT patients becomes lower than for patients receiving drugs only for the societal perspective was reached at four years.
A third pharmacoeconomic study evaluated SLIT in patients with pollen allergy and suffering from allergic rhinitis alone or in association with asthma compared with standard case-controls. [bib_ref] Comparison of costs of sublingual immunotherapy and drug treatment in grass-pollen induced..., Berto [/bib_ref] This study used a longitudinal observational database operated by a network of allergy centers. Patients were randomly assigned to SLIT (plus drugs as needed) or to treatment with drugs alone. The outcome measures included use of drugs, SLIT, visits, and tests. Costs were assessed from the Italian National Health Service perspective. The results showed that the overall per patient yearly cost of treatment was higher in SLIT patients than in those who received drugs alone, both in the whole sample (€311 versus €180 per patient, respectively), and in the rhinitis (€288 versus €116) and rhinitis associated with asthma (€362 versus €230) subgroups. Patients with both rhinitis and asthma generated more costs than those with rhinitis alone in both groups. Nevertheless, considerable savings were made in terms of the cost of drugs used to treat symptoms (-22% for rhinitis, -34% for rhinitis and asthma) in SLIT patients. Previous efficacy studies [bib_ref] Sublingual immunotherapy for allergic rhinitis: Systematic review and meta-analysis, Wilson [/bib_ref] have shown that SLIT can reduce the use of drugs for symptomatic relief, but this is the earliest research demonstrating this outcome in a patient population seen in routine clinical practice during the first year of treatment.
These studies indicate that SLIT may be beneficial to the health care system, in that it could bring more clinical effectiveness at a reduced cost in comparison with standard drug treatment, as well as have the potential to accrue extra benefit at an acceptable additional cost. 31
## Safety and tolerability
SLIT was developed with the specific goal of improving safety and this issue is of particular importance. As demonstrated in Importantly, no fatality has ever been described in association with SLIT. Most local side effects are described as mild and self-resolving without dose adjustment or medical intervention. SLIT must be discontinued very rarely. Systemic side effects (eg, asthma, rhinitis, urticaria, angioedema) have been reported only occasionally in the available controlled studies.
In 2000, André et al examined the safety of SLIT in the eight trials performed with Staloral at that time. [bib_ref] Safety of sublingual swallow immunotherapy in children and adults, André [/bib_ref] These studies involved 690 subjects (347 on active treatment and 343 on placebo), including 218 children aged 5-16 years (103 active and 115 placebo). The majority of adverse events were mild and their occurrence did not differ between the active and placebo groups. The only differences were for oral and gastrointestinal side effects that were more frequent in SLIT patients, although mild. Adverse event and dropout rates were similar in adults and children.
In 2005 a systematic review of the relevant published studies evaluated dose-related adverse event rates, expressed as the ratio between SLIT and the equivalent SCIT. [bib_ref] The safety of sublingual-swallow immunotherapy: An analysis of published studies, Gidaro [/bib_ref] In total, 25 studies were analyzed, of which 13 studies had used low doses (doses ranging from one to 50 times the dose commonly administered with SCIT), and 12 had used high doses (doses ranging from 50 to 500 times the dose administered with SCIT). Local reactions were significantly more frequent in the low-dose group than in the high-dose group (P , 0.0001), but there was no difference in systemic reaction rates between low-and highdose groups. No severe systemic reactions were reported. This review provided evidence that the occurrence of side effects during SLIT, unlike with SCIT, is not dose-dependent.
Very recently, two cases of anaphylaxis to an alternative grass pollen tablet were reported, 34 along with two respiratory reactions to the grass pollen tablet manufactured Stallergenes, [bib_ref] Sublingual immunotherapy is not always a safe alternative to subcutaneous immunotherapy, Cochard [/bib_ref] all occurring in patients who had experienced previous systemic reactions to SCIT. These reports suggest that poor tolerance of SCIT may need to be considered as a potential contraindication to use of SLIT. [bib_ref] Do indications to sublingual immunotherapy need to be revised?, Incorvaia [/bib_ref] The question of use of high-dose SLIT in children younger than five years (which is the age cut-off for SCIT) has also been addressed. An observational safety study has been done in 65 children treated with SLIT using an uptitration phase of 11 days and culminating in a maximal maintenance dose of 300 IR three times a week. [bib_ref] Safety of SLIT in children aged 3 to 7 years, Fiocchi [/bib_ref] Grass pollen was the second most frequently used allergen (after house dust mites). Adverse event rates and changes in the treatment schedule were compared in children younger or older than five years. The average cumulative dose of SLIT was 36,900 IR. Adverse reactions were observed in 11 children, none of which were severe enough to require discontinuation of immunotherapy. Six adverse reactions occurred in children younger than five years and seven in the older children. This difference was not considered significant.
## Mechanism of action
The prophylactic and therapeutic effects of immunotherapy are related to its mechanism of action. [bib_ref] Mechanisms of allergen-specific immunotherapy, Akdis [/bib_ref] In particular, immunotherapy reduces allergic inflammation even after its discontinuation, and so modifies the natural history of the allergy. Such anti-inflammatory effects, exerted also by SLIT, are based on the ability of immunotherapy to modify the phenotype of T-cells, which in allergic subjects is characterized by a prevalence of the Th2 type, with production of IL (interleukin)-4, IL-5, IL-13, IL-17, and IL-32 cytokines. [bib_ref] Mechanisms of allergen-specific immunotherapy, Akdis [/bib_ref] The induced changes result in a Th1-type response (immune deviation) related to increased IFN-gamma and IL-2 production or to reduced Th2 activity, through a mechanism of anergy or tolerance. It is now known that T-cell tolerance is characterized by the generation of allergen-specific regulatory T-cells (Treg cells), which produce cytokines, such as IL-10 and transforming growth factor-beta, with immunosuppressant and/or immunoregulatory activity. [bib_ref] Mechanisms of immune suppression by interleukin-10 and transforming growth factor beta: The..., Taylor [/bib_ref] Recent studies suggest that the antinflammatory mechanism of SLIT is similar to that of classical, subcutaneous IT, [bib_ref] On mechanism of action of sublingual immunotherapy, Incorvaia [/bib_ref] with mucosal dendritic cells having a prominent role in SLIT. Of note, it has been found in in vitro experiments that internalization of allergens by Langerhanslike dendritic cells is dose-dependent. [bib_ref] Langerhans-like dendritic cells generated from cord blood progenitors internalize pollen allergens by..., Noirey [/bib_ref] The tolerance pattern promoted by dendritic cells and driven by Treg should account for the suppressed or reduced activity of inflammatory cells and for the isotypic switch of antibody synthesis from IgE to IgG, and especially to IgG 4 . The dose-dependence of the immunologic response has also been demonstrated in clinical studies of patients treated with Staloral 300. [bib_ref] Dose dependence of immunological response to sublingual immunotherapy, Marcucci [/bib_ref] Moreover, data obtained from biopsies indicate that the oral mucosa plays a pivotal role in inducing tolerance to the sublingually administered allergen. In fact, in subjects treated with highdose grass pollen Staloral 300 SLIT, pretreatment biopsies showed very low numbers of mast cells and eosinophils (ie, the effector cells of allergic reactivity) both in the epithelium and subepithelium layers, and insignificant changes were detected after SLIT. 41
# Conclusion
SLIT is a safe and effective therapeutic alternative to SCIT for allergen immunotherapy [fig_ref] Table 3: Advantages and disadvantages of SLiT compared with SCiTAbbreviations [/fig_ref]. Staloral is an effective treatment for rhinitis and asthma attributable to sensitization to grass pollen. It is now generally accepted that allergen doses much higher than those administered by SCIT must be used to control allergic symptoms. Such doses can be administered in view of the reassuring data on tolerability and safety of high-dose SLIT. The efficacy and safety of Staloral, as demonstrated by eight double-blind, placebo-controlled studies using doses up to 1125 times higher than those administered with SCIT, [bib_ref] Efficacy of grass pollen sublingual immunotherapy for three consecutive seasons and after..., Ott [/bib_ref] [bib_ref] A double-blind, placebo-controlled trial by the sublingual route of immunotherapy with a..., Sabbah [/bib_ref] [bib_ref] Clinical efficacy of sublingual-swallow immunotherapy: A double-blind, placebo controlled trial of a..., Clavel [/bib_ref] [bib_ref] Sublingual-swallow immunotherapy (SLIT) with a standardized five-grass-pollen extract (drops and sublingual tablets)..., Pradalier [/bib_ref] [bib_ref] Randomized controlled trial of high-dose sublingual immunotherapy to treat seasonal allergic rhinitis, Smith [/bib_ref] [bib_ref] Sublingual immunotherapy in pollen-induced seasonal rhinitis and conjunctivitis: A randomized controlled trial, Mosges [/bib_ref] [bib_ref] Sublingual immunotherapy and influence on urinary leukotrienes in seasonal pediatric allergy, Yuksel [/bib_ref] [bib_ref] Efficacy and safety of high-doses sublingual immunotherapy in ultra-rush scheme in children..., Stelmach [/bib_ref] suggests that this allergen extract is an attractive therapeutic option for patients with seasonal allergies due to grass pollen.
# Disclosure
The authors report no conflicts of interest in this work.
## Drug design, development and therapy
## Publish your work in this journal
Submit your manuscript here: http://www.dovepress.com/drug-design-development-and-therapy-journal Drug Design, Development and Therapy is an international, peerreviewed open-access journal that spans the spectrum of drug design and development through to clinical applications. Clinical outcomes, patient safety, and programs for the development and effective, safe, and sustained use of medicines are a feature of the journal, which has also been accepted for indexing on PubMed Central. The manuscript management system is completely online and includes a very quick and fair peer-review system, which is all easy to use.
[fig] Figure 1: microscopic identification of components from raw materials Comparison of the protein profile with the inhouse reference by Iso electric focusing (IEF) of sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE); Dosage of the allergen activity in comparision with the in-house reference by ELISA inhibition Dosage of major allergens Extraction in optimized conditions in Good Manufacture Processing for each extract Clarifying filtration with elimination of extractive salts and low molecular weight substances Adjustment of dosage of the allergen activity in comparison with the in-house reference Standardization in Index of Reactivity (IR) Process [/fig]
[table] Table 1: Distribution of grass species in europe [/table]
[table] Table 2: Published studies of SLiT with Staloral grass pollen [/table]
[table] Table 3: Advantages and disadvantages of SLiT compared with SCiTAbbreviations: SLiT, sublingual immunotherapy; SCiT, subcutaneous immunotherapy. [/table]
|
Osteomyelitis of Fibula Rare Case with Various Differential Diagnosis
[bib_ref] Introduction to adult posttraumatic osteomyelitis of tibia, Holtom [/bib_ref] [bib_ref] A clinical staging system of adult osteomyelitis, Cierny [/bib_ref] [bib_ref] An aggressive surgical approach to the management of chronic osteomyelitis, Eckardt [/bib_ref] [bib_ref] A clinical staging system of adult osteomyelitis, Cierny [/bib_ref]
## Diagnostic criteria for chronic osteomyelitis
Clinical signs exposed bone persistent sinus tract tissue necrosis overlying bone chronic wound overlying surgical hardware chronic wound overlying fracture, laboratory evaluation positive blood cultures elevated C-reactive protein level elevated erythrocyte sedimentation rate. Treatment of osteomyelitis depends on appropriate antibiotic therapy and often requires surgical removal of infected and necrotic tissue. Choice of antibiotic therapy should be determined by culture and susceptibility results. [bib_ref] Duration of post-surgical antibiotics in chronic osteomyelitis: Empiric or evidence-based?, Haidar [/bib_ref] [bib_ref] Clinical features, diagnostic and therapeutic approaches to haematogenous vertebral osteomyelitis, Gasbarrini [/bib_ref].
Imaging studies (e.g., plain radiography, magnetic resonance imaging , and bone scintigraphy) demonstrating contiguous soft tissue infection or bony destruction.
The treatment of chronic osteomyelitis depends on stages antibiotic therapy with surgical intervention.
## Case report
A 65-year-male presented with swelling discharging sinus from anterolateral aspect middle third of the left leg since last 3 months [fig_ref] Figure 1: Sinus tract pre-operative [/fig_ref]. Patient was operated 10 years back for Tibial plateau fracture with plate and Philips screw implant [fig_ref] Figure 2: Pre-operative. [/fig_ref] [bib_ref] Treatment of infections associated with surgical implants, Darouiche [/bib_ref]. Post-operative patient recovered well.
He was asymptomatic with a medical history of diabetic mellitus for 10 years coronary artery disease for 3 years with hypertension.
On examination swelling, redness, sinuses, scar, dilated veins inflammatory signs was present over left anterolateral middle third aspect of leg with muscle wasting. Local rise in temperature, tendernes, Bony irregularity, Thickening of bone, sinuses fixity to bone over middle of leg. In Investigation blood Complete Blood Picture WBC count increased C-reactive protein level Erythrocyte sedimentation rate Fasting Blood Sugar was raised with Pus Culture report was Stap Haemolyticus growth.
X ray of leg show Middle shaft of Fibula osteolysis sclerosis irregular, unsharp borders Sequestra (shadowing encircled by a brighter surrounding) [fig_ref] Figure 2: Pre-operative. [/fig_ref]. Sinogram-Sinus tract communicating to mid shaft fibula Tibia.
Surgical technique-Implant removal was done [fig_ref] Figure 3: Post-operative X ray [/fig_ref] with abnormal tissue abnormal bone 2-10 cm fibula viaanterolateral Incision was excised out [fig_ref] Figure 4: Intraoperative operative photographs with necrotic tissue [/fig_ref] and sends for Biopsy removal of all dead bones [bib_ref] An aggressive surgical approach to the management of chronic osteomyelitis, Eckardt [/bib_ref]. Sequestra that are retained in a cavity there was no dead space as such Excising areas of active infection along the track of sinus leave a relatively healthy bleeding bone surface [bib_ref] Osteomyelitis: The past decade, Waldvogel [/bib_ref] [bib_ref] Chronic osteomyelitis. The effect of the extent of surgical resection on infection-free..., Simpson [/bib_ref] Intraoperative very unusual presentation around whole fibula was seen. Necrotic tissuse , massive debris was there which was visible like wood pieces which was later on excised out [fig_ref] Figure 4: Intraoperative operative photographs with necrotic tissue [/fig_ref] mimic as Neoplastic Lesion [bib_ref] Malignancy in chronic osteomyelitis sinus, Sadat-Ali [/bib_ref] [bib_ref] Osteomyelitis of the femur mimicking bone tumors: A review of 10 cases, Huang [/bib_ref] , Metastasis, Fungal Osteomyelitis or Granulmatous infection.
The soft tissue and Bone pieces was send for Biopsy [fig_ref] Figure 6: Intraoperative operative photographs with necrotic tissue for biopsy [/fig_ref] and histopathology culture. Patient was given 2 weeks of Intra venous Antibiotics and 6 weeks oral antibiotics (along with Rifampicin) with DM control [bib_ref] Duration of post-surgical antibiotics in chronic osteomyelitis: Empiric or evidence-based?, Haidar [/bib_ref] [bib_ref] Osteomyelitis: The past decade, Waldvogel [/bib_ref] [bib_ref] Adjunctive use of rifampin for the treatment of Staphylococcus aureus infections: A..., Perlroth [/bib_ref]. This case is follow of 2 year with satisfactory result. Non weight bearing for 4 Weeks with Knee and Ankle Exercises later allowed for full weight bearing.
# Discussion
Osteomyelitis staging system was first described by Waldvogel in 1970 distinguished three etiologic routes secondary to hematological infection [bib_ref] Osteomyelitis: The past decade, Waldvogel [/bib_ref]. Bacteraemia is casused by distant foci of infection may lead to without clinical signs of sepsis. Changes in plain radiographs can include scalloping of the cortex and periostal reaction soft-tissue swelling, osteopenia, [bib_ref] Osteomyelitis in long bones, Lazzarini [/bib_ref]. CT scanning be helpful in the identification of sequestra while MRI is more useful for soft-tissue assessment and revealing early bony oedema [bib_ref] Osteomyelitis in long bones, Lazzarini [/bib_ref].
Treatment of Chronic Osteomyelitis include Antibiotic Mader et al. [bib_ref] A clinical staging system of adult osteomyelitis, Cierny [/bib_ref] [bib_ref] Diagnosis and management of chronic infection, Forsberg [/bib_ref]. recommend regimens based on their staging system. Type 1 osteomyelitis is treated with 4 weeks of parenteral antibiotics. Type 2 often resolves after 2 weeks of antibiotics after debridement. Types 3 and 4 each require 4 weeks of parenteral antibiotics from the last debridement. Some centers use 2 weeks of parenteral antibiotics followed by 4 weeks of oral antibiotics. Surgical management of osteomyelitis consists of two basic steps; debridement and obliteration of the subsequent dead space by soft tissue [bib_ref] Duration of post-surgical antibiotics in chronic osteomyelitis: Empiric or evidence-based?, Haidar [/bib_ref]. Adequate surgical debridement removes dead necrotic tissues, decreases the bacterial load, and gives a chance for the host immune system and antibiotics to arrest infection. Adequate debridement may leave a large bony defect, or dead space All sequestra must be removed and resection of scarred and infected bone and soft tissue is indicated [bib_ref] An aggressive surgical approach to the management of chronic osteomyelitis, Eckardt [/bib_ref] [bib_ref] Duration of post-surgical antibiotics in chronic osteomyelitis: Empiric or evidence-based?, Haidar [/bib_ref]. Periosteal stripping should be avoided because it may result in avascularity and the involucrum surrounding the infection can be left in place [bib_ref] Osteomyelitis: The past decade, Waldvogel [/bib_ref]. Debridement of the bone is performed until the 'paprika sign' is seen; a pin-point bleeding noted on the viable bone [bib_ref] Surgical management of chronic osteomyelitis, Parsons [/bib_ref]. The dead space created. Numerous surgical options can be used to treat bone defects such as primary shortening, cancellous bone graft in one stage, cancellous bone graft in two stages (after induced membrane technique), pedicle bone transfer, periosteal free flap transfer, or bone morphogenetic protein. Although non-vascularized bone grafting is effective in small bone defect with well-perfused soft tissues, they are less reliable when the gap defect is greater than 6 cm and when soft tissue vascularization is poor [bib_ref] Adult osteomyelitis, Calhoun [/bib_ref].
Osteomyelitis of the fibula is uncommon. If the patient's tibia is not involved remove a sequestrum from his fibula, without waiting for an involucrum to form, because his tibia will support body weight. Expose any part of his fibula by approaching it between his peroneal muscles anteriorly and his soleus posteriorly posterior tibial nerve and vessels are protected; but be careful not to injure his peroneal artery and veins which are close to the postero-medial angle of the shaft of his fibula. Intraoperative the dead necrotic abnormal tissue bone wood like pieces mimicked neoplastic lesion Fungal Osteomyelitis Granulmatous infection but after Biopsy it was diagnosed Chronic Osteomyelitis of Fibula with adequate debridement curettage and with proper antibiotic coverage result was satisfactory.
# Conclusion
The Osteomyelitis of fibula is clinical diagnosis with support of investigation can be presented with various differential diagnosis clinically and intraoperatively but after proper debridement curettage and biopsy report probable diagnosis is made with proper duration of antibiotic 2 weeks of Intra venous Antibiotics 6 weeks oral antibiotics (along with Rifampicin), chronic osteomyelitis is treated with satisfactory result with good outcome instead initially mimicked various differential diagnosis.
## Clinical message
Diagnosing the chronic Osteomyelitis with various differential diagnoses can be challenging if presented intr-aoperative with abnormal tissue wood like pieces. A detailed evaluation is essential. After through debridement proper coverage of antibiotic with proper duration and confirm biopsy the diagnosis is confirmed and proper line of treatment is outlined with good result. The abnormal tissues wood like pieces is very rare to see during surgery post operative outcome was very satisfactory.
[fig] Figure 1: Sinus tract pre-operative. [/fig]
[fig] Figure 2: Pre-operative. [/fig]
[fig] Figure 3: Post-operative X ray. Journal of Orthopaedic Case Reports Journal of Orthopaedic Case Reports | Volume 7 | Issue 3 | May -Jun 2017 | Page 59-62 [/fig]
[fig] Figure 4: Intraoperative operative photographs with necrotic tissue. [/fig]
[fig] Figure 7: (a and b) Post-operative after 6 months. [/fig]
[fig] Figure 5: Intraoperative operative photographs with abnormal necrotic tissue. [/fig]
[fig] Figure 6: Intraoperative operative photographs with necrotic tissue for biopsy. a b [/fig]
|
New Politics, an Opportunity for Maternal Health Advancement in Eastern Myanmar: An Integrative Review
# Introduction
Myanmar (formerly Burma) is a populous and impoverished southeast Asian country, with a long history of military rule and political conflicts characterized by widespread human rights violations. Myanmar's population of pregnant women are particularly at risk, and that risk is significantly higher among pregnant women of ethnic minority groups in the eastern provinces. Very limited data for the strife-ridden eastern provinces have made it difficult to establish a pattern of the prevalence of maternal mortality in Myanmar. Officially, the country's Central Statistical Organization reported the nation's urban maternal mortality ratio (MMR) to be 123 per 100,000 livebirths and the rural MMR to be 157 (1). International bodies, such as the WHO, UNICEF, UNFPA, and World Bank, however, suggest that the overall MMR may be as high while nominally free, the reality is that many health services require a user fee (9) that the Government has imposed as a so-called community cost-sharing scheme (10). Family planning services, such as contraception, are difficult to obtain, particularly in the eastern regions (8). As recently as 2000, family planning services were virtually unavailable in According to the Ministry of Health, the national unmet need for family planning is 17.7% (11), a rate similar to that found by the UN (12). However, the unmet need in the eastern regions has been documented at greater than 60% (13).
Many women in Myanmar resort to illegal abortions as the primary means of contraception . Officially, Myanmar makes antenatal care available to 83% of pregnant women nationally (5); however, other estimates from the UN suggest that the national rate may be closer to 76% (12). In the eastern regions, Mullany and colleagues found that, prior to any intervention on their part, antenatal care in that area covered less than 40% of pregnant women (13). Even with antenatal care coverage, many mothers still suffer from poor nutrition, anaemia, or malaria. A significant lack of skilled birth attendants has been noted; one midwife is often responsible for five to 16 villages (15).
The Ministry of Health, quoting data from 2008, suggested that 76% of pregnant women received skilled attendance (11). Other non-governmental bodies suggested that it may be closer to . suggested that as many as 73% of Myanmar's women gave birth in village homes in the care of nonprofessionals (17). In contrast, skilled attendance in the eastern regions has been documented as low as 5% (13). Emergency obstetric care is inadequate nationally (5).
Accessing care is particularly difficult in the eastern provinces since the few available hospitals are at great distances from villages (18) and are dangerous to access due to landmines and ongoing conflict between the military and ethnic militias, seeking greater autonomy or independence for their regions. Ethnicity further places women at risk since the Myanmar Government consistently appears to provide fewer services to ethnic minorities. Healthcare is more deficient in regions of ethnic minority groups , and infrastructure is particularly poor in the rural areas of these regions (20).
Myanmar has struggled under military rule for more than 50 years. The military consumes 40% of the national budget while merely 2.2% is allo-cated to healthcare (5). The long-standing military government in Myanmar was also well-known for human rights abuses revolving around ongoing conflict with ethnic minorities (13). Research by the BPHWT has linked poor health indicators to the human rights abuses experienced in the eastern provinces . It is in this context that recent political liberalization in Myanmar has been encouraging. In March 2011, a nominal civilian government, headed by President Thein Sein, was elected, representing Myanmar's first non-military rule since 1962. The new government has sought to demonstrate leniency, recently releasing over 2100 political detainees (21). Among those released from house arrest was Aung San Suu Kyi, an internationally-renowned democratic figurehead. Finally, in recent years, the Myanmar Government has signed ceasefire treaties with several ethnic rebel groups (22).
The principal aim of this paper is to carry out an integrative review of the literature on the current status of maternal mortality in eastern Myanmar in the context of armed conflict between various separatist groups and the military regime. In the process, underlying factors contributing to maternal mortality in eastern Myanmar are examined. The authors discuss possible opportunities that liberalization may make available to the Myanmar Government, the international community, and non-governmental organizations (NGOs) in the fight to reduce maternal mortality.
# Materials and methods
## Literature search
This integrative review of literature sought qualitative, quantitative and unpublished studies, and grey literature (grey literature refers to documents of the Ministry of Health, organizational reports, country documents, book chapters, and newspaper articles). An exhaustive electronic search, in English, through scholarly databases (Medline and Embase) and full-text journal databases (Proquest, Ovid and Science Direct) was conducted using the following key words: Burma, Myanmar, childbirth, civil war, death, development agencies, international aid, government, liberalization, maternal deaths, maternal health, maternal mortality, MMR, political unrest, policy, and pregnancy. The corresponding authors of publications in the field of maternal mortality in Myanmar were contacted via a standardized email for any outstanding unpublished articles. Of those authors who responded, only one had authored additional material on the subject. Grey literature was examined with Google and Google Scholar search engines, using the same search terms mentioned above. Unpublished documents were sought in electronic theses libraries. Hand-searching of relevant journals was also completed for the period spanning the six months prior to 31 July 2012 to ensure that any recent publications not yet available in databases were not overlooked. No relevant documents were found. All publications that entered the second phase of screening (see below) were examined for relevant titles from their references lists.
## Screening of the studies
Documents selected for the review included published journal articles (n=8) and grey literature (n=10) published in English from 1 January 2004 to 31 July 2012. English was found to be the main language of these studies done in this area, and limited translation resources did not allow us to include any non-English publications. The dates chosen reflected the eight years prior to the study in order to reflect a recent picture of the maternal health situation in Myanmar (very few studies exist to further limit the time span and still retain a variety of information sources). Studies on refugees on the Thai-Myanmar border areas were excluded in order to prevent differing service levels between regions that would affect the results. An initial screening was carried out by examining titles and abstracts for relevance. Searching by key words occasionally generated papers with entirely unrelated and different contents, and these were excluded from the review. A second-stage screening was completed on full-length publications which met the above criteria. Document selection was conducted by the primary author in consultation with the tertiary author.
## Data extraction
Data from qualitative and quantitative studies and grey literature were extracted thematically via a line-by-line coding process (23) and entered in a spreadsheet [fig_ref] Table 1: Data-collection tool Author [/fig_ref]. Study designs and outcomes from quantitative studies were also extracted.
# Data analysis
Qualitative data were coded line-by-line, and then categorized into descriptive themes (23). The themes were then regrouped into four further analytical themes based on principal aim and objectives of this study-a process based on the work of . These four themes were underlying factors of MMR in the eastern regions 1a is not missing, it is the first item in the 'MHC Services' section of the table. The items are not in sequence because, when these were grouped into broader themes (as described in the methods section), these maintained their numbering and lettering
## Contd.
relating to: the health system, conflict, politics, and socioeconomics. The quantitative data were coded into the descriptive themes described above and translated into the analytical themes to seek sources of convergence (25). Quantitative data were presented in a narrative style as well as numerically, to address the objective of each study. Statistical metaanalysis was beyond the scope of this study, and there were only a few quantitative studies to meet the requirements for such an analysis. Findings from the quantitative studies helped complement the themes emerging from the qualitative studies.
## Data synthesis
Analytical themes were generated based upon the combined qualitative and quantitative findings to synthesize the findings into an overarching framework (23). Some themes were deductively created (e.g. political factors) where the authors postulated broader themes likely to be present in the literature while others evolved through an inductive process where specific findings in the literature were generalized into a broader theme (e.g. socioeconomic factors). Trust necessary for timely care to be achieved Some EOC services can be delivered in the community Challenges to MHC service delivery were overcome (e.g. security problems, transportation issues) Successes: impartial care and consistent support of communities in turmoil, increased MHW confidence, expansion of family planning services, increased collaboration between 3 tiers (Traditional birth attendant, health worker, maternal health worker), expansion of TBA roles in some communities (e.g. prophylactic misoprostol)
Contd. [fig_ref] Table 2: Study summary [/fig_ref] provides a summary of the quantitative (n=5) and qualitative studies (n=2) and review articles (n=1) found through a detailed literature search. [fig_ref] Table 3: Extracted themes [/fig_ref] lists the 11 themes, which were synthesized from these studies (n=8) in combination with a qualitative summary study (n=1) and publications sourced via the grey literature search (n=10). Underlying factors that influenced high MMR in eastern Myanmar could be broadly categorized into the following themes:
# Results
## Health system-related underlying factors for mmr in the eastern regions
Access to and availability of maternal healthcare (MHC) services was an underlying factor identified in the review. It appeared in 11 of the 19 publications studied. found that, despite a maternal mortality rate higher than in other southeast Asian countries, per-capita aid-flows to Myanmar were among the lowest in the region (17). They described the health system as being on the verge of collapse in 2012. A report by the Hauser Center suggested that short-term donor funding prevented adequate healthcare service delivery by the participating NGOs in the study (26). For instance, 88% of women delivered their babies at home (13).
In a 2008 study, the authors stated that it was necessary to decentralize the provision of healthcare to properly serve the eastern regions of Myanmar where a community-based delivery system could allow many of the necessary pregnancy and delivery services to be provided (27). The three-tier health worker system used by certain NGOs in the area was designed whereby maternal health workers provided a wide range of MHC services, and health workers and traditional birth attendants (TBAs) provided a narrower range of services (27). This three-tier system has improved the accessibility of care for many women in the region (28). However, the overall lack of availability and of userfriendliness of MHC services was compounded by difficult access (transport, fees, etc.) (13,28). (13). In a later study, Mullany and colleagues found that, after implementing a community-based MHC project, women were more likely to receive antenatal care (71.8% vs 39.2%), were twice as likely to receive postnatal care, and were 10 times more likely to have skilled attendance at birth (28). In a qualitative study based upon the same project (29), the authors found that some EOC services can be delivered in a community setting, particularly with increased collaboration between the three tiers of service providers and an expansion of the role of TBAs in certain regions.
Contraception was identified as an area of concern by four of the publications. Approximately 80% of women in two BPHWT studies had not used contraception . A study by found that greater than 60% of women had an unmet need for contraception (13). In the eastern regions, older women reported an average of 6.9 pregnancies, with 88% of their most recent births occurring at home (13). Forced relocation was associated with 6.1 times decreased use of contraception (8).
The MOM project implemented community-based MHC, which resulted in a decrease in unmet contraceptive need from 61.7% to 40.5% of the women surveyed, and an increased use in modern contraception from 23.9% to 45% (28).
## Conflict-related underlying factors for mmr in the eastern regions
Conflict-related factors are difficult to link directly to maternal mortality. However, these factors play Removal of food and the means to purchase food Forced labour Forced relocation or displacement Targeting of medical staff, buildings, and supplies Health system-related underlying factors of MMR in the eastern regions Access to and availability of MHC* services Presence of MHC services Contraception Underlying political factors of MMR in the eastern regions Limitations placed upon NGOs The lack of health-related data Socioeconomic underlying factors of MMR in the eastern regions Co-existing morbidity among pregnant women Personal violation or violence was identified as a concern in the majority (15/19) of the publications considered. Experience of serious forms of violence was widespread; 0.6% of households in the eastern region reported experiencing a gunshot injury, and another 1.7% experienced a landmine injury in the year preceding one study. In an earlier study by the same group, 1.9% of households studied reported experiencing physical assault or violence perpetrated by military personnel (8
## Underlying political factors of mmr in the eastern regions
Government-imposed limitations on NGOs working in Myanmar and the hindrances the organizations face were underlined as areas of concern in 11 of the 19 reviewed publications. The Hauser Center's publication identified a number of findings, including a lack of expatriate staff mobility, fluctuating visa approvals, and governmentimposed travel restrictions (26). These findings are compounded by limited NGOs' access to regions of ethnic minority groups within Myanmar and serious ethical and operational dilemmas. Identified ethical dilemmas included how to provide appropriate assistance in a humanitarian environment that is tightly controlled by the Government and the military. The creation of parallel structures (through aid-delivery by multiple sources) was found to be another significant operational dilemma facing NGOs in according to a report by.
A lack of health-related data in Myanmar and in the eastern regions in particular (including maternal health data) was identified as an underlying politically-driven factor by 5 of the 19 publications. For instance, NGOs have difficulty in gathering data on the impact of their services due to the restricted mobility of their expatriate staff (26). Myanmar's Ministry of Health regularly underestimates national health indicator data (8), and these data are generalized for the eastern regions where, in fact, minimal data are collected. Therefore, the unique health needs in the eastern regions are neglected.
## Underlying socioeconomic factors of mmr in the eastern regions
Co-existing morbidity among pregnant women in eastern Myanmar was identified as an underlying factor in 5 of 19 publications. In the BPHWT study, 7.3% the of the household heads were found to be positive for falciparum malaria (4), a figure closely mirrored by another study which found a rate of 7.2% in the women surveyed (13). Bednets were present in 21.6% of households (13). Malaria was the largest cause of death in another study (8).
Eighteen percent of all those surveyed in BPHWT's 2010 study were malnourished (4) while another study noted that many of those surveyed were malnourished with an average mid-upper armcircumference of 24.4 cm among adult females (13). The same study (13) found that over 50% of women surveyed were anaemic with haemoglobin counts of less than 11 g/dL in greater than 60%.
Only 11.8% of women had received iron supplements during their previous pregnancy (13) while, in another study, only 14.7% of women met international recommendations for iron supplementation. An earlier study found that 40% had received iron supplementation (8). The same BPHWT study found that greater than 60% rarely or never used a latrine, and greater than 30% rarely or never boiled their water (8). Diarrhoea was a common finding in the studies, with rates of 6.4% in the two weeks prior to the survey (4) and 9.8% in the households surveyed by the BPHWT (8).
Abortion was not reported or documented as a factor of importance to MMR in any of the publications reviewed. The studies indicate that the factors that appear to contribute to the higher maternal mortality observed in eastern Myanmar are strongly interconnected.
## Figure. web of underlying causes of maternal mortality in eastern myanmar
# Discussion
## Interrelationship of underlying factors influencing mmr
The many inter-relationships among the underlying factors which contribute to maternal mortality in Myanmar's eastern regions are posited in the figure. The model in the figure was created to illustrate the complex inter-relationships of the themes that emerged from the review. The number of interrelationships is complex, and the authors discuss a selected few in the next section.
## Health system-related underlying factors for mmr in the eastern regions
Access to and availability of MHC services present serious challenges to pregnant women due, in part, to their lack of availability from both Government and NGOs. Travelling to access services increased the likelihood of exposure to soldiers, which risks violence upon their persons, or exposure to landmines. Targeted health workers are also presumably less likely to risk confronting the military in order to ensure the availability of MHC services.
## Conflict-related underlying factors of mmr in the eastern regions
The alarmingly high rates of exposure to HRVs, such as personal violation or violence which includes sexual violence can have a direct impact on the health of pregnant women. For instance, violence perpetrated by soldiers may place a pregnancy in jeopardy. Women who conceive as a result of these encounters may be less likely to access formal maternal healthcare (MHC) services as was the case in Rwanda (31). Even those who did not conceive might still be less likely to access MHC services when they did conceive, due to shame ensuing from their victimization, as suggested in research on victims of rape in northern Uganda (32). Furthermore, anecdotal evidence from Sierra Leone suggested that pregnancies secondary to conflict-related sexual violence may increase abortion-seeking, including unsafe abortions (33).
The removal of food as well as the means to purchase food may directly impact the health of pregnant women in eastern Myanmar since one study found it to be linked with an increased risk of death (30). Furthermore, women foraging for food in the jungle are more likely to come into contact with malaria-carrying mosquitoes (30), soldiers, or landmines. Women injured by soldiers were also less likely to be able to contribute to the raising of crops, which might impact on food security.
Forced labour prevented households from growing their own food or working in a remunerated capacity. Pregnant women forced into such labour were also more likely to be exposed to sexual or physical violence due to their close proximity to soldiers. Forced labour conditions were often unsanitary and unfamiliar, further exposing people to landmines, malaria (30), and other diseases. These conditions also reduced their access to appropriate food sources.
Forced relocation and displacement had a serious impact on food security and nutritional status of pregnant women since households were relocated away from fields and food storage facilities. It also increased women's risk of contracting malaria (e.g. not sleeping under a bednet), a condition which claimed the lives of many displaced pregnant women during the Afghan conflict (34). The odds of exposure to landmines were also increased with forced displacement (8) since women travelled in dangerous areas and foraged foods in unfamiliar territory.
## Underlying political factors of mmr in the eastern regions
The lack of health-related data, particularly in the area of maternal health, from the eastern regions of Myanmar, is closely related to the difficulty that NGOs have in accessing this region. A lack of data on the severity of health concerns affected programme planning (26) and, in turn, the availability of the funding of NGOs. Ongoing conflict made it difficult to conduct research (35) due, in part, to the dangers associated with data collection in the region. A lack of appropriate data also posed challenges for the Myanmar Government and other service-delivery agents to making appropriate decisions regarding which MHC services were most needed. Limited data impacted on planning (26) which contributed to the poorly-tailored and geographically-disparate MHC service provision.
## Emerging political changes
Myanmar has made encouraging changes recently; by-elections won by Aung San Suu Kyi's National League for Democracy and changes to the constitution and political structures suggest a liberalization and democratization of the country. The Government has also softened its attitude towards NGOs by giving them greater access to certain regions within the country (26). Additionally, the Human Rights Commission has been formed, and a number of military members have recently been sentenced for raping and murdering a woman of an ethnic minority group . Finally, a recent drive to produce a national ceasefire with the majority of Myanmar's ethnic armed groups is underway (38). These events all point to the possibility of substantial and real change in Myanmar. However, much remains that is troubling. Human rights violations continue (39)
## Limitations of the review
The present review was conducted using the available publications that remain limited despite a wide-ranging literature search using a thorough, methodical and replicable process. Publication bias, typical of such reviews (43), was countered through the collection of unpublished documents, such as dissertations and grey literature. The quality of data may also be suspected, based on the data variability from source to source, i.e. Ministry of Health (8).
A subset of the publications included in the study focused on the reporting of HRVs in eastern Myanmar where maternal health concerns were a secondary focus or were not directly addressed.
Although the methodological rigour of the studies examined were practical for the conflict-prone areas under consideration, this did not allow us to determine the relative weights of the actual causes of increased maternal mortality in eastern Myanmar. A similar cohort of women receiving services from either the BPHWT or the MOM group (those who conducted the studies in question) could be included in a subsequent study to improve the strength of the results found. It is difficult to make a one-to-one link between some of the conflictrelated factors and maternal death from these studies. Nevertheless, these factors can all contribute to an environment that is detrimental to the healthy progress of pregnancy or maternal survival.
# Conclusions
Further progress in Myanmar is essential to reduce maternal mortality in the eastern regions. To reduce the conflict-related underlying causes, measures must be taken to reduce the human rights violations occurring in the eastern regions. The Myanmar Government must develop a comprehensive plan that truly engages the ethnic minority groups in serious dialogue (22), one that can lead to national reconciliation. A call has been put forth by Suu Kyi and other political figures for holding a second Panglong Conference to bring ethnic leaders, together with the Government (39). This could be an important first step.
Furthermore, the Myanmar Government should enact laws which enshrine the rights of pregnant women. Without changing the prevalence of these conflict-related underlying causes (found in the study to be of greatest concern to women's health), the authors suggest that few other measures will have the lasting effect needed to reduce MMR in the eastern regions of Myanmar.
The conflict-related barriers to accessing MHC services are ongoing but can be mitigated. The provision of the government-funded mobile maternal health services, as suggested by , to the eastern regions, would greatly reduce these barriers. An alternative might be the creation of numerous health "safe zones" where ethnic women could access MHC services free from harassment or molestation by the military. This approach should be complemented by further funding and support from the international community for the existing exemplary cross-border initiatives that currently provide MHC services to this population.
Health system-related underlying factors contributing to high maternal mortality in Myanmar's eastern regions must also be addressed: a greater focus on maternal health provision to the communities in eastern Myanmar is crucial to reducing the unnecessary deaths (13,27). Another essential element of MHC that must be expanded is the provision of family planning services, or, "birth spacing services" in the terms of the Ministry of Health. Contraceptive services could be provided by government-employed community health workers who travel to the villages in which the minority women live. Finally, the underlying socioeconomic factors must not be ignored. Aid and humanitarian agencies must partner with the Myanmar Government, increasing aid-flows desperately needed to combat high maternal mortality in the eastern regions (17).
[table] Table 1: Data-collection tool Author .................................................................... Year ......................................................................... Title ......................................................................... [/table]
[table] Table 2: Study summary [/table]
[table] Table 3: Extracted themes: influences on maternal mortality in eastern Myanmar [/table]
|
Incorporating Distant Sequence Features and Radial Basis Function Networks to Identify Ubiquitin Conjugation Sites
Ubiquitin (Ub) is a small protein that consists of 76 amino acids about 8.5 kDa. In ubiquitin conjugation, the ubiquitin is majorly conjugated on the lysine residue of protein by Ub-ligating (E3) enzymes. Three major enzymes participate in ubiquitin conjugation. They are -E1, E2 and E3 which are responsible for activating, conjugating and ligating ubiquitin, respectively. Ubiquitin conjugation in eukaryotes is an important mechanism of the proteasome-mediated degradation of a protein and regulating the activity of transcription factors. Motivated by the importance of ubiquitin conjugation in biological processes, this investigation develops a method, UbSite, which uses utilizes an efficient radial basis function (RBF) network to identify protein ubiquitin conjugation (ubiquitylation) sites. This work not only investigates the amino acid composition but also the structural characteristics, physicochemical properties, and evolutionary information of amino acids around ubiquitylation (Ub) sites. With reference to the pathway of ubiquitin conjugation, the substrate sites for E3 recognition, which are distant from ubiquitylation sites, are investigated. The measurement of F-score in a large window size (220,+20) revealed a statistically significant amino acid composition and position-specific scoring matrix (evolutionary information), which are mainly located distant from Ub sites. The distant information can be used effectively to differentiate Ub sites from non-Ub sites. As determined by five-fold cross-validation, the model that was trained using the combination of amino acid composition and evolutionary information performs best in identifying ubiquitin conjugation sites. The prediction sensitivity, specificity, and accuracy are 65.5%, 74.8%, and 74.5%, respectively. Although the amino acid sequences around the ubiquitin conjugation sites do not contain conserved motifs, the cross-validation result indicates that the integration of distant sequence features of Ub sites can improve predictive performance. Additionally, the independent test demonstrates that the proposed method can outperform other ubiquitylation prediction tools.
# Introduction
Ubiquitin (Ub) is a small protein that consists of 76 amino acids about 8.5 kDa. Ubiquitin conjugation sites of protein (Ubiquitylation), which is an essential post-translational modification, is a sequential process that involves a group of enzymes known as E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase). The ubiquitylation system is well-known for the selective degradation of serveral short-lived proteins in eukaryotic cells [bib_ref] The ubiquitin system, Hershko [/bib_ref]. The attachment of a ubiquitin or poly-ubiquitin chains to proteins influences serveral cellular processes, including transcriptional regulation, signal transduction, development, apoptosis, endocytosis, and cell proliferation [bib_ref] Control of protein degradation by E3 ubiquitin ligases in Drosophila eye development, Ou [/bib_ref]. Ubiquitin is mostly conjugated on the lysine residue of a protein by Ub-ligating (E3) enzymes. The E3 ligase must be sufficiently specific and must act only on a defined subset of cellular targets to ensure signal fidelity [bib_ref] Ubiquitin-binding domains, Hicke [/bib_ref]. Another enzyme, E4, has that can stabilize and extend a poly-ubiquitin chain, has also been found [bib_ref] Degradation signals for ubiquitin system proteolysis in Saccharomyces cerevisiae, Gilon [/bib_ref].
With the development of high-throughput tandem mass spectrometry-based proteomics, the number of studies of the comprehensive identification ubiquitylated proteins and their conjugated sites is increasing [bib_ref] A proteomics approach to understanding protein ubiquitination, Peng [/bib_ref]. UbiProt [bib_ref] UbiProt: a database of ubiquitylated proteins, Chernorudskiy [/bib_ref] identified all experimentally verified ubiquitin-conjugated sites from the publicly literature. Some of the entries include information about enzyme data obtained by enzyme purification and isolation. The entries supply annotations of the ubiquitin-conjugated and the exact positions of the ubiquitin-conjugation sites. UniProtKB/ Swiss-Prot [bib_ref] The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003, Boeckmann [/bib_ref] is a comprehensively annotated protein database. Both experimentally validated and putative ubiquitin-conjugated annotations can be obtained from the post-translation modification annotations in the database.
Experimental identifications of ubiquitin-conjugation sites on ubiquitylated proteins in vivo and in vitro are the foundation for understanding the mechanisms of ubiquitination dynamics. However, these experiments are commonly time-consuming, labor-intensive and expensive. the in silico prediction of ubiquitin-conjugated sites with high predictive performance could be promising for preliminary analyses and could greatly reduce the number of potential targets that require further in vivo or in vitro confirmation. UbiPred [bib_ref] Computational identification of ubiquitylation sites from protein sequences, Tung [/bib_ref] used an algorithm for mining informative physicochemical properties from protein sequences to train SVM-based ubiquitylation site prediction system. Based on leave-one-out cross-validation, the SVM model that is trained with 31 physicochemical properties was evaluated. It was found to improve the predictive accuracy from 72.1% to 84.4%. Recently, Radivojac et al. [bib_ref] Identification, analysis, and prediction of protein ubiquitination sites, Radivojac [/bib_ref] have investigated that the sequence biases and structural preferences around known ubiquitination sites are similar to those of intrinsically disordered protein regions. Additionally, Radivojac et al. developed a random forest predictor of ubiquitination sites, UbPred, that could reach a balanced accuracy of 72%.
Given the importance of ubiquitin conjugation in biological processes, this investigation presents a method, UbSite, in which an efficient radial basis function (RBF) network is utilized to identify protein ubiquitin conjugation (ubiquitylation) sites. The experimentally verified ubiquitylated proteins and ubiquitylation sites are collected from UbiProt [bib_ref] UbiProt: a database of ubiquitylated proteins, Chernorudskiy [/bib_ref] and UniProtKB/Swiss-Prot [bib_ref] The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003, Boeckmann [/bib_ref]. Not only amino acid composition, but also structural characteristics, physicochemical properties, and evolutionary information of amino acids around the ubiquitylation (Ub) sites are investigated. With reference to the pathway of ubiquitin conjugation, which is by sequential process that involves a group of enzymes, E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase), the distant sequence features of ubiquitylation sites for E3 recognition are investigated. A position specific scoring matrix (PSSM), which is generated by PSI-BLAST [bib_ref] Gapped BLAST and PSI-BLAST: a new generation of protein database search programs, Altschul [/bib_ref] search against a non-redundant database of protein sequences, is utilized to study the evolutionary information surround the ubiquitin conjugation sites. The constructed PSSM is regarded as a measure of residue conservation in a window of a given length. Based on the measurement of F-score in a large window size (220,+20), the statistically significant amino acid composition and evolutionary information, which are mainly located at positions distant from the ubiquitylation sites, can be utilized effectively to differentiate ubiquitylation sites from non-ubiquitylation sites. An evaluation of the trained models based on five-fold cross-validation revealed, that the prediction sensitivity, specificity and accuracy were 65.5%, 74.8%, and 74.5%, respectively. The independent test demonstrates shows that UbSite outperform previous ubiquitylation prediction tools.
# Materials and methods
As shown in [fig_ref] Figure 1: The analytic flowchart of UbSite [/fig_ref] , the proposed approach, UbSite, is composed of three major analytical steps -data collection and preprocessing, feature extraction, and model training and evaluation. This investigation comprehensively analyzes the structural characteristics and physicochemical properties that surround the ubiquitin conjugation sites. The details of the analysis are described as follows.
## Data collection and preprocessing
Experimentally confirmed ubiquitin conjugation sites are collected from UbiProt [bib_ref] UbiProt: a database of ubiquitylated proteins, Chernorudskiy [/bib_ref] and UniProtKB/Swiss-Prot [bib_ref] The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003, Boeckmann [/bib_ref]. UbiProt consists of 158 experimentally confirmed ubiquitinconjugation sites. Then, we extracted the sequences from release 57.0 of UniProtKB/Swiss-Prot if the sequences are annotated as 'ubiquitin' in the 'MOD_RES' fields. We also removed the sites that are annotated as ''by similarity'', ''potential'' or ''probable''. A total of 337 entries are annotated as ubiquitin-conjugated proteins in UniProtKB/Swiss-Prot, and they include 416 ubiquitylation sites. After removing the redundant data from UbiProt and UniProtKB/Swiss-Prot, a total of 442 experimental ubiquitylation sites associated with 350 ubiquitylated proteins are obtained. In this work, the 442 experimental ubiquitylation sites are regarded as the positive dataset.
To prevent overestimation of the predictive performance, homologous sequences are removed from the training data by using a window size of 2n+1 for ubiquitylation sites. With reference to the reduction process in MASA [bib_ref] Incorporating structural characteristics for identification of protein methylation sites, Shien [/bib_ref] , two ubiquitylated protein sequences with more than 30% identity were defined as homologous sequences. Then, two homologous sequences were specified to re-align the fragment sequences using a window length of 2n+1, centered on the ubiquitylation sites using BL2SEQ [bib_ref] BLAST 2 Sequences, a new tool for comparing protein and nucleotide sequences, Tatusova [/bib_ref]. For two fragment sequences with 100% identity, when the ubiquitylation sites in the two proteins are in the same positions, only one site was kept. The homologous negative data were also reduced by using the same approach.
With respect to classification, the predictive performance of the trained models may be overestimated because of the over-fitting of a training set. To estimate the real predictive performance, the experimental ubiquitylation sites, whose annotated dates are after , are selected as the independent test set. As shown in [fig_ref] Table 1: The statistics of non-homologous training data and independent test data for ubiquitylation... [/fig_ref] , the data in the non-homologous training set include 385 positive sites (ubiquitylation) and 12582 negative sites (nonubiquitylation) in 301 ubiquitylated proteins. The data of the non-homologous independent test set include 57 positive sites and 3502 negative sites in 49 ubiquitylated proteins. Following the evaluation by five-fold cross-validation, the trained model with the highest accuracy was further evaluated based on independent test data. The independent test sets were utilized to test not only the proposed method but also the previously proposed ubiquitylation prediction tools, UbiPred [bib_ref] Computational identification of ubiquitylation sites from protein sequences, Tung [/bib_ref] and UbPred [bib_ref] Identification, analysis, and prediction of protein ubiquitination sites, Radivojac [/bib_ref].
## Feature extraction and coding
Coding of amino acid sequences. Fragments of amino acids are extracted from positive and negative training sets using a window of length 2n+1 that is centered on ubiquitylation sites. Various values of n are used to determine the optimal window length. The BLOSUM62 matrix is adopted to represent the protein primary sequence information as the basic feature set for learning radial basis function networks. A matrix of m6n elements is used to represent each residue in a training dataset, where n denotes the window size and m = 21, which elements comprise 20 amino acids and one terminal signal. Each row of the normalized BLOSUM62 matrix is adopted to encode one of 20 amino acids.
Compositions of amino acids and amino acid pairs. A total of n vectors {x i , i = 1, …, n} were used, to represent all n proteins in the training data. Each vector is labeled with the group of its corresponding protein (e.g. ubiquitylated or nonubiquitylated). The vector x i has 20 elements for the amino acid composition and 400 elements for the amino acid pair composition. The 20 elements specify the numbers of occurrences of 20 amino acids normalized with the total number of residues in the protein, and the 400 elements specify the numbers of occurrences of 400 amino acid pairs normalized with the total number of residues in the protein. In this investigation, amino acid composition and amino acid pair composition are combined, yielding, 420 elements in each vector.
Position Specific Scoring Matrix Profiles. In the point of view of structure, several amino acid residues of a protein can be mutated without changing its structure, and two proteins may have similar structures with different amino acid compositions. Position Specific Scoring Matrix (PSSM) profiles, which have been extensively utilized in protein secondary structure prediction, subcellular localization and other bioinformatics problems are adopted herein with significant improvement [bib_ref] Protein secondary structure prediction based on positionspecific scoring matrices, Jones [/bib_ref] [bib_ref] LOCSVMPSI: a web server for subcellular localization of eukaryotic proteins using SVM..., Xie [/bib_ref] [bib_ref] TMBETADISC-RBF: Discrimination of beta-barrel membrane proteins using RBF networks and PSSM profiles, Ou [/bib_ref]. The PSSM profiles were obtained by PSI-BLAST against nonredundant fragment sequences of Ub sites. [fig_ref] Figure 2: The detailed process of generating position specific scoring matrix [/fig_ref] displays in detail how to generate the 400D PSSM features for each ubiquitylation and non-ubiquitylation site. The matrix of m620 elements has rows centered on ubiquitylation or non-ubiquitylation site, extracted from the PSSM profile, where m represents the window size and 20 represents the position specific scores for each type of amino acid. Thereafter, the m620 matrix is transformed into a 20620 matrix by summing up the rows that are associated with the same type of amino acid. Finally, every element in 20620 matrix is divided by the window length m and then is normalized using the formula: 1 1ze {x . Structural characteristics. Since most of the experimental ubiquitylated proteins do not have corresponding protein tertiary structures in PDB [bib_ref] The Protein Data Bank, Berman [/bib_ref] , an effective tool, RVP-Net [bib_ref] RVP-net: online prediction of real valued accessible surface area of proteins from..., Ahmad [/bib_ref] , was used to compute the ASA value based on the protein sequence. The computed ASA value is the percentage area of each amino acid on the proteins that is accessible to the solvent. RVP-net applies a neural network to predict real ASA values of the residues based on neighborhood information, with a mean absolute error of 18.0-19.5%, defined as the absolute difference between the predicted and experimental values of relative ASA per residue [bib_ref] RVP-net: online prediction of real valued accessible surface area of proteins from..., Ahmad [/bib_ref]. Fulllength protein sequences with experimental ubiquitylated sites are input to RVP-Net to compute the ASA value for all residues. The ASA values of amino acids that surround the ubiquitylation site were extracted and scaled from zero to one.
PSIPRED [bib_ref] The PSIPRED protein structure prediction server, Mcguffin [/bib_ref] was utilized to compute the secondary structure that surrounds the ubiquitylation sites from the protein sequence. PSIPRED is a simple and reliable method for predicting secondary structure, which applies two feed-forward neural networks to analyze the output obtained from PSI-BLAST [bib_ref] Gapped BLAST and PSI-BLAST: a new generation of protein database search programs, Altschul [/bib_ref]. PSIPRED 2.0 achieved a mean Q 3 score of 80.6% across all 40 submitted target domains without obvious sequence similarity with structures that are present in PDB; accordingly, PSIPRED has been ranked as the best of 20 evaluated methods [bib_ref] Protein structure prediction servers at University College London, Bryson [/bib_ref]. The output of PSIPRED is ''H,'' ''E'' or ''C'', which stand for helix, sheet and coil, respectively. The full-length protein sequences with ubiquitylation sites are inputted to PSIPRED to determine the secondary structure of all residues. The orthogonal binary coding approach is adopted to transform the three terms that specify the secondary structure into numeric vectors. For instance, helix is encoded ''100;'' sheet is encoded ''010,'' and coil is encoded ''001.'' F-score measurement. To study further the specificity of the substrate sites, the features that statistically differ between ubiquitylation sites and non-ubiquitylation sites are identified, based on a statistical measurement of F-score. The F-score of the ith feature is defined as, k,i denotes the ith feature of the kth negative instance. Fscore supports a simple approach for measuring features that are more discriminative. If the i-th feature has a high F-score, then this feature effectively discriminates between positive and negative datasets.
[formula] F À score i ð Þ~x z ð Þ i {x i 2 z x { ð Þ i {x i 2 1 n z {1 X n z k~1 x z ð Þ k,i {x z ð Þ i 2 z 1 n { {1 X n { k~1 x { ð Þ k,i {x { ð Þ i 2ð1Þ [/formula]
## Training and evaluation of model
In this work, the QuickRBF package [bib_ref] A novel radial basis function network classifier with centers set by hierarchical..., Ou [/bib_ref] has been employed to construct radial basis function network (RBFN) classifiers. As presented in [fig_ref] Figure 1: The analytic flowchart of UbSite [/fig_ref] (See Supplementary Materials), the general architecture in an RBFN consists of three layers, namely the input layer, the hidden layer, and the output layer. The input layer broadcasts the coordinates of the input vector to each of the nodes in the hidden layer. Each node in the hidden layer then produces an activation based on the associated radial basis kernel function. Finally, each node in the output layer computes a linear combination of the activations of the hidden nodes. The general mathematical form of the output nodes in RBFN is as follows:
[formula] c j x ð Þ~X k i~1 w ji w x{m i k k; s i ð Þ ;ð2Þ [/formula]
where c j x ð Þ denotes the function corresponding to the j-th output node and is a linear combination of k radial basis functions w() with center m i and bandwidth s i ; Also, w ji denotes the weight associated with the correlation between the j-th output node and the i-th hidden node. In this work, we adopted a fixed bandwidth (s) of five, and used all input nodes as centers (k = n). With its several bioinformatics applications, classification based on radial basis function network has been extensively adopted to predict factors such as the cleavage sites in proteins [bib_ref] Bio-basis function neural network for prediction of protease cleavage sites in proteins, Yang [/bib_ref] , inter-residue contacts [bib_ref] Prediction of inter-residue contacts map based on genetic algorithm optimized radial basis..., Zhang [/bib_ref] , protein disorder [bib_ref] Protein disorder prediction by condensed PSSM considering propensity for order or disorder, Su [/bib_ref] , discrimination of b-barrel proteins [bib_ref] TMBETADISC-RBF: Discrimination of beta-barrel membrane proteins using RBF networks and PSSM profiles, Ou [/bib_ref] , and identification of O-linked glycosylation [bib_ref] Incorporating significant amino acid pairs to identify O-linked glycosylation sites on transmembrane..., Chen [/bib_ref].
Predictive performance of the constructed RBFN classifier is evaluated by performing k-fold cross validation. The original data (training data in [fig_ref] Table 1: The statistics of non-homologous training data and independent test data for ubiquitylation... [/fig_ref] is divided into k subgroups by splitting each dataset into k approximately equal sized subgroups. In one round of cross-validation, a subgroup is regarded as the test set, and the remaining k-1 subgroups are regarded as the training set. The cross-validation process is repeated k rounds, with each of k subgroups used as the test set in turn. Then, the k results are combined to produce a single estimation. The advantage of k-fold cross-validation is that all original data are regarded as both training set and test set, and each data is used for test exactly once [bib_ref] A study of cross-validation and bootstrap for accuracy estimation and model selection, Ron [/bib_ref].
# Results and discussion
## Amino acid composition of ubiquitin conjugation sites
This investigation focuses on the analysis of ubiquitin conjugated lysine. In ubiquitin conjugation, the region of the ubiquitinconjugated lysine residues is in directly contact with the E3 ligase catalytic center. Since E3 ubiquitin ligase enzymes have a substrate binding specificity, whether the region of ubiquitinconjugated lysine conserved amino acid motifs for E3 ubiquitin ligase recognition must be explored. After the duplicated sequences of experimental ubiquitylation sites are removed, as shown in [fig_ref] Figure 3: The position-specific amino acid composition, accessible surface area and secondary structure of... [/fig_ref] , the amino acids composition that flanked the ubiquitin-conjugated lysines (ubiquitylation site centered at position 0) are graphically visualized as a 41-mer sequence logo.
WebLogo [bib_ref] WebLogo: a sequence logo generator, Crooks [/bib_ref] [bib_ref] Sequence logos: a new way to display consensus sequences, Schneider [/bib_ref] is adapted to generate the graphical sequence logo for the relative frequency of the amino acid at each position around the ubiquitylated sites. The conservation of amino acids around the ubiquitylation sites can then be easily investigated. Based on the sequence logo representation, the most abundant residues of the ubiquitylation sites are the charged and polar amino acids, including Aspartic acid (D) and Glutamic acid (E). The amino acids around the modified sites are not obviously conserved, a slight difference between the preferences of amino acids for the ubiquitylation and non-ubiquitylation sites.
Since the representation of sequence logos involves different preferences of amino acids for ubiquitylated and non-ubiquitylated sites, the statistical difference in the distribution of amino acids around ubiquitylated (Ub) and non-ubiquitylated (non-Ub) lysines is calculated. [fig_ref] Figure 2: The detailed process of generating position specific scoring matrix [/fig_ref] (See Supplementary Materials) displays the compositional differences between Ub and non-Ub sites. The more abundant amino acids at the Ub sites are Alanine (A), Aspartic acid (D), Glycine (G) and Isoleucine (I), and the depleted hydrophobic residues around these sites include Cysteine (C) and Leucine (L) around Ub sites. Moreover, the Lysine (K) and Serine (S) are less abundant around Ub sites. The amino acid sequences around the ubiquitin-conjugated sites can be alternatively grouped by various methods to generalize the sequence feature because amino acid classification is hierarchical. As presented in [fig_ref] Table 1: The statistics of non-homologous training data and independent test data for ubiquitylation... [/fig_ref] (See Supplementary Materials), the three-class grouping method and the eight-class grouping method are used to 20 amino acids into subgroups that capture their chemical properties. Three-class grouping methods can be based on hydrophobicity [bib_ref] The classification and origins of protein folding patterns, Chothia [/bib_ref] , polarity [bib_ref] Amino acid difference formula to help explain protein evolution, Grantham [/bib_ref] , normalized van der Waals volume [bib_ref] Amino acid side chain parameters for correlation studies in biology and pharmacology, Fauchere [/bib_ref] and polarizability [bib_ref] The structural dependence of amino acid hydrophobicity parameters, Charton [/bib_ref]. Additionally, a Two Sample Logo [bib_ref] Two Sample Logo: a graphical representation of the differences between two sets..., Vacic [/bib_ref] of 41-mer compositional biases around Ub conjugation sites compared to non-Ub conjugation sites is presented in [fig_ref] Figure 3: The position-specific amino acid composition, accessible surface area and secondary structure of... [/fig_ref] (See Supplementary Materials). The amino acid residues that significantly enriched and depleted (P-value ,0.05; t-test) around Ub conjugation sites are shown. With the investigation of position-specific difference of amino acid composition in 41-mer window length, [fig_ref] Figure 3: The position-specific amino acid composition, accessible surface area and secondary structure of... [/fig_ref] indicates the positions that are distant from Ub sites have statistically significant differences of amino acid composition.
## Structural characteristics of ubiquitin conjugation sites
A side-chain of an amino acid that undergoes post-translational modification preferentially accesses the surface of a protein [bib_ref] Surface accessibility of protein posttranslational modifications, Pang [/bib_ref]. To investigate the preference of the solvent accessible surface area [bib_ref] SeqVISTA: a new module of integrated computational tools for studying transcriptional regulation, Hu [/bib_ref] that surrounds ubiquitin conjugation sites in protein tertiary structures, the experimentally identified ubiquitylation sites are mapped to the corresponding positions of the protein entries in the Protein Data Bank (PDB) [bib_ref] The Protein Data Bank, Berman [/bib_ref]. The preference of the secondary structure around the ubiquitylation sites is also considered. Since most of the experimentally confirmed ubiquitylated proteins do not have corresponding protein tertiary structures in PDB [bib_ref] The Protein Data Bank, Berman [/bib_ref] , RVP-Net [bib_ref] RVP-net: online prediction of real valued accessible surface area of proteins from..., Ahmad [/bib_ref] and PSIPRED [bib_ref] The PSIPRED protein structure prediction server, Mcguffin [/bib_ref] are adopted to compute the ASA value and secondary structure, respectively, from the protein sequence. [fig_ref] Figure 3: The position-specific amino acid composition, accessible surface area and secondary structure of... [/fig_ref] presents the sequence logo of the secondary structure and the average percentage of ASA in the 41-mer window (220,+20) of the ubiquitylation (Ub) and non-ubiquitylation (non-Ub) sites. In the investigation of secondary structure around the Ub sites, Catic et al. [bib_ref] Preferred in vivo ubiquitination sites, Catic [/bib_ref] has found the preference for coil structure. In this work, the observations reveal that Ub ligase (E3) prefers to recognize the regions that are located in coil (loop) or helix structures. In contrast to Ub sites, non-Ub sites don not have an obviously preferred secondary structure. In the study of solvent accessibility, most of the Ub or non-Ub lysines are located in the highly accessible surface area. However, the mean solventaccessible surface area that surrounds the Ub sites slightly exceeds that around non-Ub sites.
## Investigation of distant sequence features for ubiquitylation sites
Owing to the direct interaction between the enzyme and the substrate site, most of the proposed PTM prediction methods investigate amino acid sequences close to the modified sites. The ubiquitin conjugation pathway, which involves a sequential process with a group of enzymes known as E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase) [bib_ref] The ubiquitin system, Hershko [/bib_ref] motivates an investigation of the distant sequence features that are distant from ubiquitylation sites. In [fig_ref] Figure 4: The hypothetic model of identifying the distant sequence features for E3 recognition [/fig_ref] , a graphical model represents the hypothesis that contains a substrate site that is distant from ubiquitylated lysine for E3 recognition. Ubiquitin is mostly conjugated on the lysine residue of a protein by substrate recognition of Ub-ligating (E3) enzymes [bib_ref] Control of protein degradation by E3 ubiquitin ligases in Drosophila eye development, Ou [/bib_ref]. E3 enzymes function as the substrate recognition modules of the system and are capable of interaction with both E2 and substrate. Thus, the E3 ligase must be sufficiently specific and must act only on a defined subset of cellular targets to ensure signal fidelity [bib_ref] Ubiquitin-binding domains, Hicke [/bib_ref]. Based on the measurement of the F-score in a large window size (220,+20), displays the statistically significant composition of amino acids at positions 216, 210, 23, 21, +1, +5, +10, +13, and +17. The surrounding positions that have high F-scores are (significant for differentiating the ubiquitylation sites from the non-ubiquitylation sites. Additionally, Tables S2 and S3 (See Supplementary Materials) present the significant amino acids and di-peptides in the surrounding region (220,+20), which have a higher F-score value.
Position specific scoring matrix (PSSM), which is generated by PSI-BLAST [bib_ref] Gapped BLAST and PSI-BLAST: a new generation of protein database search programs, Altschul [/bib_ref] search against a non-redundant database of protein sequence, is utilized to obtain evolutionary information about amino acids around the ubiquitin conjugation sites. The constructed PSSM includes the probability that each amino acid is present at each position. Therefore, PSSM is regarded as a measure of residue conservation in a window of a particular length. displays statistically significant evolutionary information concerning amino acids at each position in the window from 220 to +20. Based on the measurement of F-scores, the positions 219, 217, 215, 212, 210, 24, 21, +5, +9, +13, +15 and +18, where the F-scores are highest, are significant for differentiating the ubiquitylation sites from the non-ubiquitylation sites. In the investigation of distant sequence features that are distant from ubiquitylation sites, the length of the training data window for learning the predictive model is set to 41-mer (220,+20).
To demonstrate the distant sequence features are informative for the identification of ubiquitylation sites, herein, five-fold crossvalidation is performed to evaluate the models trained with various window sizes 2n+1, where n varies from five to twenty. The predictive models (RBFN classifiers) are trained with the feature of amino acid composition. [fig_ref] Figure 7: The predictive performance of the models trained with different window length varying... [/fig_ref] presents the sensitivity, specificity, and accuracy of cross-validation based on various window lengths.
As different window sizes from 11-mer to 41-mer are applied, the predictive accuracy improves from 63.1% to 73.7%, the sensitivity and specificity increase as well. Especially for the window length which is longer than 35-mer, the predictive power is apparently improved with the accuracy that is higher than 70.0%. As the investigation of distant sequence features in , the model that was trained with a large window length performs better than that without the distant sequence features.
## Predictive performance of cross-validation using various training features
Most predictive models are based on the features of amino acid sequences. To determine which features can be utilized to construct . The statistically significant composition of amino acids for each position in the window length from 220 to +20. Based on the measurement of F-score, the positions 216, 210, 23, 21, +1, +5, +10, +13, and +17, containing higher value of F-score, are significant for differentiating the ubiquitylation sites from non-ubiquitylation sites. doi:10.1371/journal.pone.0017331.g005 . The statistically significant evolutionary information of amino acids for each position in the window length from 220 to +20. Based on the measurement of F-score, the positions 219, 217, 215, 212, 210, 24, 21, +5, +9, +13, +15 and +18, containing higher value of Fscore, are significant for differentiating the ubiquitylation sites from non-ubiquitylation sites. doi:10.1371/journal.pone.0017331.g006 models that differentiate between ubiquitylation sites and nonubiquitylation sites, various features, including the sequence of amino acids, amino acid composition, accessible surface area, and physicochemical properties are evaluated by k-fold cross-validation. The amino acids (AA) and accessible surface area (ASA) around the ubiquitylated sites are encoded using a BLOSUM62 matrix and the RVP-Net-computed ASA values, respectively. [fig_ref] Table 2: The predictive performance of cross-validation using various training features [/fig_ref] presents the predictive performance achieved using various training features, based on five-fold cross-validation. Of the models trained using individual features, those that are trained using amino acid composition slightly outperform those that are trained using amino acids, ASA, the secondary structure, or PSSMs. In particular, the model trained with the PSSM profile of non-redundant ubiquitylated protein sequences achieves an accuracy of 70%. However, the model that is trained with the secondary structure underperforms prediction based on ubiquitylation sites. According to the F-score of distant sequence features, the amino acid composition and evolutionary information (PSSM) at several flanking positions are statistically differently distributed between ubiquitylation sites and non-ubiquitylation sites. Therefore, the effects of combining informative features are evaluated. As presented in [fig_ref] Table 2: The predictive performance of cross-validation using various training features [/fig_ref] , the model that is trained using the combination of amino acid composition and PSSM profile of non-redundant fragment sequences of Ub sites performs best, with the best-balanced predictive sensitivity and specificity.
## Predictive performance of independent testing
To determine whether the models (are over-fitted to their training data, independent sets of data concerning Ub sites and non-Ub sites are constructed and used to test the model that was trained with the combination of amino acid composition and the PSSM profile of non-redundant fragment sequences of Ub sites, which have the highest predictive accuracy. Independent test sets include 57 ubiquitylation sites and 3502 non-ubiquitylation sites, According to [fig_ref] Table 3: Comparison between our method [/fig_ref] , are used to determine the predictive sensitivity, specificity, and accuracy of the proposed method, which were 57.9%, 72.4%, and 72.2%, respectively. Generally, the performance in an independent test approaches that of cross-validation. Whereas crossvalidation outperforms independent testing, the performance of the trained model may be overestimated. The independent test establishes that the constructed RBF model does not over-fit the training data. The independent test sets were used to test other ubiquitylation predictors. The predictive sensitivity and specificity of UbiPred [bib_ref] Computational identification of ubiquitylation sites from protein sequences, Tung [/bib_ref] were 52.6% and 52.6%, respectively, indicating balanced predictive performance. However, the independent test also indicates that UbiPred does not perform as well as its developers claimed. The predictive sensitivity and specificity of UbPred [bib_ref] Identification, analysis, and prediction of protein ubiquitination sites, Radivojac [/bib_ref] were 42.1% and 68.7%, respectively, indicating poor sensitivity for an independent test set. In UbiPred and UbPred, the data source of experimental ubiquitylation sites is collected from UbiProt [bib_ref] UbiProt: a database of ubiquitylated proteins, Chernorudskiy [/bib_ref] , which mainly stored the yeast ubiquitylation data. However, UbSite integrates the experimental ubiquitylation sites from UbiProt and UniProtKB/Swiss-Prot [bib_ref] The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003, Boeckmann [/bib_ref] , which accumulated the ubiquitylation data from multiple species. This could partially explain the low sensitivities of UbPred and UbiPred on these independent test data which come from multiple species.
# Conclusion
This investigation proposes a method, UbSite, which incorporates the efficient radial basis function (RBF) network to identify ubiquitin conjugation sites on protein sequences. Not only the amino acid composition but also the structural characteristics, physicochemical properties, and evolutionary information of amino acids around the ubiquitylation (Ub) sites are explored. With reference to the pathway of ubiquitin conjugation, which involves a sequential process with a group of enzymes known as E1 (activating enzyme), E2 (conjugating enzyme) and E3 (ubiquitin ligase), the substrate sites for E3 recognition, which are distant from ubiquitylation sites, are examined. According to the measurement of F-score in a large window size(220,+20), most of the statistically significant amino acids and evolutionary information (PSSM), which can be used effectively to differentiate Ub sites from non-Ub sites, are located at large distances from Ub sites. To prevent any overestimation of predictive performance, duplicated sequences are removed using a window size determined by the collected data sets. Although the amino acid sequences around the ubiquitin conjugated sites do not contain a conserved motif, cross-validation results indicate that the integration of the evolutionary information around the sites can improve predictive performance. [fig_ref] Table 3: Comparison between our method [/fig_ref] compares proposed method with other ubiquitylation prediction tools, in terms of materials, method, training features, number of training data, window length, and proposed predictive performances. Furthermore, the independent test demonstrates that UbSite can outperform other ubiquitylation prediction tools.
Although the proposed method is accurate and robust, according to independent tests, some issues remain to address in the future. Firstly, the structural preferences of ubiquitin conjugation sites preferred structures at ubiquitin conjugation sites must be examined in greater detail because flanking residues are not conserved. In addition to the solvent-accessible surface area and the secondary structure, the B-factor, intrinsically disordered region, protein linker region, and other factors should be explored experimentally at ubiquitylation sites in the protein regions with PDB entries. Following work done previously on phosphorylation [bib_ref] Phospho3D: a database of three-dimensional structures of protein phosphorylation sites, Zanzoni [/bib_ref] , the local 3D structure of ubiquitylation sites may be extracted for further analysis. Secondly, with reference to the pathway of ubiquitin conjugation, the ubiquitylated proteins may contain a substrate site is distant from ubiquitylation lysine and useful in E3 recognition. Therefore, the distant sequence features of ubiquitylation sites should be investigated in more detail. For instance, using the motif discovery tool, like MEME [bib_ref] MEME: discovering and analyzing DNA and protein sequence motifs, Bailey [/bib_ref] , to explore the significant motifs which may be the substrate sites of E3 recognition. [fig_ref] Figure 1: The analytic flowchart of UbSite [/fig_ref] The general architecture of RBFN consisting of input layer, hidden layer, and output layer.
## Supporting information
(TIF) [fig_ref] Figure 2: The detailed process of generating position specific scoring matrix [/fig_ref] The compositional differences of amino acids around ubiquitylation sites compared to non-ubiquitylation sites. (TIF) [fig_ref] Figure 3: The position-specific amino acid composition, accessible surface area and secondary structure of... [/fig_ref] A Two Sample Logo of the compositional biases around Ub conjugation sites compared to non-Ub conjugation sites. The amino acid residues that significantly enriched and depleted (P-value,0.05; t-test) around Ub conjugation sites are shown.
(TIF) [fig_ref] Table 1: The statistics of non-homologous training data and independent test data for ubiquitylation... [/fig_ref] The graphical representation of chemical properties surrounding ubiquitylation sites using different grouping method.
(DOC)
[fig] Figure 1: The analytic flowchart of UbSite. doi:10.1371/journal.pone.0017331.g001 [/fig]
[fig] Figure 3: The position-specific amino acid composition, accessible surface area and secondary structure of ubiquitin conjugated lysines and non-ubiquitin conjugated lysines. doi:10.1371/journal.pone.0017331.g003 [/fig]
[fig] Figure 2: The detailed process of generating position specific scoring matrix (PSSM) and encoding the fragment of amino acid sequence by generated PSSM. doi:10.1371/journal.pone.0017331.g002 [/fig]
[fig] Figure 4: The hypothetic model of identifying the distant sequence features for E3 recognition. doi:10.1371/journal.pone.0017331.g004 [/fig]
[fig] Figure 7: The predictive performance of the models trained with different window length varying from 11-mer to 41-mer. doi:10.1371/journal.pone.0017331.g007 [/fig]
[table] Table 1: The statistics of non-homologous training data and independent test data for ubiquitylation and non-ubiquitylation sites. [/table]
[table] Table 2: The predictive performance of cross-validation using various training features. [/table]
[table] Table 3: Comparison between our method (UbSite) and other ubiquitylation prediction tools. Abbreviation: AAC, amino acid composition; PSSM, position-specific scoring matrix. doi:10.1371/journal.pone.0017331.t003 [/table]
[table] Table S2 F: -score of amino acid composition for 40 positions around Ubi site. (DOC) [/table]
[table] Table S3: Top 20 di-peptides with high value of F-score in the 41-mer window size (220,+20) around Ub site. (DOC) [/table]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.